Electrochemical Detection of Arsenic in Various Water Samples
Electrochemical Arsenic Remediation for Rural Bangladesh by ...
Transcript of Electrochemical Arsenic Remediation for Rural Bangladesh by ...
Electrochemical Arsenic Remediation for Rural Bangladesh
by
Susan Elizabeth Amrose Addy
B.S. (University of Michigan at Ann Arbor) 2000M.A. (University of California at Berkeley) 2003
A dissertation submitted in partial satisfaction of the
requirements for the degree of
Doctor of Philosophy
in
Physics
in the
GRADUATE DIVISION
of the
UNIVERSITY OF CALIFORNIA, BERKELEY
Committee in charge:Professor Ashok Gadgil, Co-Chair
Professor Robert Jacobsen, Co-ChairProfessor Richard MullerProfessor David Sedlak
Fall 2008
The dissertation of Susan Elizabeth Amrose Addy is approved:
Co-Chair Date
Co-Chair Date
Date
Date
University of California, Berkeley
Electrochemical Arsenic Remediation for Rural Bangladesh
Copyright 2008
by
Susan Elizabeth Amrose Addy
1
Abstract
Electrochemical Arsenic Remediation for Rural Bangladesh
by
Susan Elizabeth Amrose Addy
Doctor of Philosophy in Physics
University of California, Berkeley
Professor Ashok Gadgil, Co-Chair
Professor Robert Jacobsen, Co-Chair
Arsenic in drinking water is a major public health problem threatening the lives
of over 140 million people worldwide. In Bangladesh alone, up to 57 million people
drink arsenic-laden water from shallow wells. ElectroChemical Arsenic Remediation
(ECAR) overcomes many of the obstacles that plague current technologies and can be
used affordably and on a small-scale, allowing for rapid dissemination into Bangladesh
to address this arsenic crisis.
In this work, ECAR was shown to effectively reduce 550 - 580 µg/L arsenic (in-
cluding both As[III] and As[V] in a 1:1 ratio) to below the WHO recommended maximum
limit of 10 µg/L in synthetic Bangladesh groundwater containing relevant concentrations
of competitive ions such as phosphate, silicate, and bicarbonate. Arsenic removal ca-
pacity was found to be approximately constant within certain ranges of current density,
but was found to change substantially between ranges. In order of decreasing arsenic
2
removal capacity, the pattern was: 0.02 mA/cm2 > 0.07 mA/cm2 > 0.30 - 1.1 mA/cm2
> 5.0 - 100 mA/cm2. Current processing time was found to effect arsenic removal ca-
pacity independent of either charge density or current density. Electrode polarization
studies showed no passivation of the electrode in the tested range (up to current density
10 mA/cm2) and ruled out oxygen evolution as the cause of decreasing removal capacity
with current density. Simple settling and decantation required approximately 3 days to
achieve arsenic removal comparable to filtration with a 0.1 µm membrane.
X-ray Absorption Spectroscopy (XAS) showed that (1) there is no significant
difference in the arsenic removal mechanism of ECAR during operation at different cur-
rent densities and (2) the arsenic removal mechanism in ECAR is consistent with ar-
senate adsorption onto a homogenous Fe(III)oxyhydroxide similar in structure to 2-line
ferrihydrite.
ECAR effectively reduced high arsenic concentrations (100 - 500 µg/L) in real
Bangladesh tube well water collected from three regions to below the WHO limit of 10
µg/L. Prototype fabrication and field testing are currently underway.
Professor Ashok GadgilDissertation Committee Co-Chair
Professor Robert JacobsenDissertation Committee Co-Chair
i
This work is dedicated
to my husband,
Nathan Joseph Addy
my love, my heart, and my partner in all things,
to my father,
Fredrick James Amrose
who has encouraged me and supported me for my entire life,
and to the friend I will always admire most,
Maya Sophia James
who has saved my life and my sanity and helped shape my outlook on all things.
ii
Contents
List of Figures vi
List of Tables xii
1 Introduction 11.1 Arsenic contamination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Worldwide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.2 Crisis in Bangladesh and West Bengal . . . . . . . . . . . . . . . . 4
1.2 Arsenic remediation and safe water sources . . . . . . . . . . . . . . . . . 61.3 A new implementation model for Bangladesh . . . . . . . . . . . . . . . . 71.4 ElectroChemical Arsenic Remediation (ECAR) . . . . . . . . . . . . . . . 10
1.4.1 ECAR operating parameters . . . . . . . . . . . . . . . . . . . . . 121.4.2 Other considerations - treatment time and cost . . . . . . . . . . . 131.4.3 Previous electrocoagulation research . . . . . . . . . . . . . . . . . 141.4.4 The need for parameter studies in Bangladesh groundwater . . . . 16
1.5 Thesis objectives and dissertation structure . . . . . . . . . . . . . . . . . 171.5.1 Chapter 2: Relevant Scientific Background . . . . . . . . . . . . . 181.5.2 Chapter 3: Chemical and Physical Analysis of Arsenic Complexa-
tion with Iron in ECAR . . . . . . . . . . . . . . . . . . . . . . . . 181.5.3 Chapter 4: Characterization of Reaction Products . . . . . . . . . 201.5.4 Chapter 5: ECAR Performance in Real Bangladesh Groundwater . 21
2 Relevant Scientific Background 232.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.2 Arsenic in Natural Waters . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.3 Iron (Hydr)oxides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.4 Arsenic sorption onto iron (hydr)oxides . . . . . . . . . . . . . . . . . . . 30
2.4.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302.4.2 Review of arsenic adsorption literature . . . . . . . . . . . . . . . . 352.4.3 Arsenic removal via zero valent iron . . . . . . . . . . . . . . . . . 382.4.4 Effect of pH and speciation on arsenic sorption . . . . . . . . . . . 40
iii
2.4.5 Kinetics of arsenic sorption . . . . . . . . . . . . . . . . . . . . . . 432.4.6 Effects of co-occuring solutes . . . . . . . . . . . . . . . . . . . . . 48
2.5 Electrocoagulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542.5.1 Electrochemical cell theory . . . . . . . . . . . . . . . . . . . . . . 542.5.2 Electrocoagulation (EC) using iron . . . . . . . . . . . . . . . . . . 622.5.3 As[III] oxidation in EC . . . . . . . . . . . . . . . . . . . . . . . . 732.5.4 Arsenic removal mechanism using EC . . . . . . . . . . . . . . . . 752.5.5 Arsenic removal studies using EC . . . . . . . . . . . . . . . . . . . 752.5.6 Kinetics of EC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 782.5.7 Characterization of EC products . . . . . . . . . . . . . . . . . . . 79
2.6 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3 Chemical and Physical Analysis of Arsenic Complexation with Iron inECAR 843.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.1.1 Operating parameters in ECAR . . . . . . . . . . . . . . . . . . . . 843.1.2 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 903.2.1 Arsenic analysis and arsenic speciation . . . . . . . . . . . . . . . . 903.2.2 pH, dissolved oxygen measurements . . . . . . . . . . . . . . . . . 923.2.3 Development of Synthetic Bangladesh Groundwater (SBGW) . . . 933.2.4 Electrochemical cell . . . . . . . . . . . . . . . . . . . . . . . . . . 1013.2.5 Electrochemical Arsenic Remediation Procedure . . . . . . . . . . 1053.2.6 ECAR batch experiments run 1 - Matrix 1 . . . . . . . . . . . . . 1083.2.7 ECAR batch experiments run 2 - Matrix 2 . . . . . . . . . . . . . 1173.2.8 Adsorption using post-synthesis ECAR-generated adsorbent . . . . 1223.2.9 Polarization studies . . . . . . . . . . . . . . . . . . . . . . . . . . 1243.2.10 Electrochemical Impedance Spectroscopy . . . . . . . . . . . . . . 1283.2.11 Sedimentation tests . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333.3.1 Resistivity of synthetic Bangladesh water . . . . . . . . . . . . . . 1333.3.2 Arsenic removal capability in synthetic Bangladesh water . . . . . 1363.3.3 Effect of charge density . . . . . . . . . . . . . . . . . . . . . . . . 1433.3.4 Effect of current density . . . . . . . . . . . . . . . . . . . . . . . . 1463.3.5 Effect of current processing time . . . . . . . . . . . . . . . . . . . 1523.3.6 Effect of phosphate and silicate on arsenic removal . . . . . . . . . 1543.3.7 Parameter trends, tradeoffs, and implications for Bangladesh . . . 1573.3.8 Adsorption using post-synthesis ECAR-generated adsorbent . . . . 1583.3.9 As[III] removal in ECAR . . . . . . . . . . . . . . . . . . . . . . . 1643.3.10 Polarization studies . . . . . . . . . . . . . . . . . . . . . . . . . . 1693.3.11 Sedimentation versus 0.1µm vacuum filtration . . . . . . . . . . . . 175
3.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
iv
4 Characterization of Reaction Products 1894.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
4.1.1 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 1924.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
4.2.1 Arsenic breakthrough as a function of membrane pore size . . . . . 1944.2.2 Scanning Electron Microscopy (SEM) . . . . . . . . . . . . . . . . 1944.2.3 Chemical analysis of EGA sludge . . . . . . . . . . . . . . . . . . . 1954.2.4 X-ray Absorption Spectroscopy (XAS) . . . . . . . . . . . . . . . . 196
4.2.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 1964.2.4.2 Iron XANES and XRF-maps . . . . . . . . . . . . . . . . 1984.2.4.3 Null arsenic EXAFS attempts . . . . . . . . . . . . . . . 2024.2.4.4 Arsenic k-edge EXAFS . . . . . . . . . . . . . . . . . . . 2044.2.4.5 Iron k-edge EXAFS . . . . . . . . . . . . . . . . . . . . . 2064.2.4.6 Reference spectra . . . . . . . . . . . . . . . . . . . . . . . 2074.2.4.7 XAS data analysis . . . . . . . . . . . . . . . . . . . . . . 208
4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2094.3.1 Limit on the average EGA cluster size . . . . . . . . . . . . . . . . 2094.3.2 Scanning Electron Microscopy (SEM) . . . . . . . . . . . . . . . . 2104.3.3 XRF Fe-As mapping of EGA sludge . . . . . . . . . . . . . . . . . 2144.3.4 Chemical analysis of waste EGA sludge . . . . . . . . . . . . . . . 2154.3.5 Iron oxidation state in EGA . . . . . . . . . . . . . . . . . . . . . . 2184.3.6 EGA iron structure - comparison to known iron (hydr)oxides . . . 2194.3.7 Arsenic bonding structure - comparison to known arsenic-iron com-
plexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2224.3.8 EGA iron structure - comparison between current densities . . . . 2244.3.9 Arsenic bonding structure - comparison between current densities . 2264.3.10 Arsenic capture in the bulk solution compared to the electrode
surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2274.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
5 ECAR Performance in Real Bangladesh Groundwater 2315.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
5.1.1 Issues of groundwater transport and storage . . . . . . . . . . . . . 2325.1.2 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 234
5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2355.2.1 Arsenic analysis and arsenic speciation . . . . . . . . . . . . . . . . 2355.2.2 Tube well water sample collection and storage . . . . . . . . . . . . 2355.2.3 Measurements of As[III]/Astot ratio over time . . . . . . . . . . . . 2395.2.4 Chemical analysis for co-occuring solutes . . . . . . . . . . . . . . 2415.2.5 ECAR Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2455.3.1 As[III] monitoring in samples of real groundwater . . . . . . . . . . 2455.3.2 ECAR performance in real groundwater . . . . . . . . . . . . . . . 2495.3.3 Mixing time in real groundwater compared to synthetic groundwater252
v
5.3.4 Sample well comparison to BGS - co-occurring solutes . . . . . . . 2565.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
6 Conclusions 2656.1 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2656.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
A List of Acronyms 292
vi
List of Figures
1.1 A schematic detailing the roles, cash flow, and mutual benefit relationshipsbetween each of the partnering institutions in a public-private partnership(PPP). The PPP is the key to long term viability of a community cleanwater center. Appropriate safe water technology would be licensed to theprivate company. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 Schematic representation of the removal of arsenate ions from solutionvia EC. For this schematic, Fe 3+ ions are dissolving from the electrode,though Fe 2+ ions may dissolve as well. In addition, Fe(OH)3 is used torepresent the precipitating iron (hydr)oxides, though the actual composi-tion of iron (hydr)oxides formed may be quite complex (see text). . . . . . 68
2.2 Pourbaix diagram for iron. Assumes iron is the only species present.Eh is the equilibrium potential relative to a standard hydrogen electrode(adapted from Hem (1961)) . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.1 The ratio of As[III] to total-arsenic (Astot) over a period of 2-3 days forthree representative water batches (WB) of the SBGW-1 type. For waterbatches 38 and 45, no initial (0 day) measurement of As[III] was done, soaverage As[III]/Astot ratio of monitored water batches (8 total) at day-0was used with the standard deviation as an error. . . . . . . . . . . . . . . 101
3.2 Schematic representation of 850 mL-separated cathode electrochemical cell.1033.3 Schematic representation of 3 L-single chamber electrochemical cell. . . . 1043.4 Fraction of remaining arsenic as a function of mixing time for SBGW-
1. Both kinetics tests were performed under identical conditions after adosing at j = 1.1 mA/cm2, q = 39 C/L, tCP = 5 min, with initial arsenicAstot ≈ 1000 µg/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
vii
3.5 Replicate variability between arsenic concentrations measured in identicalbatch tests (repeat measurements) for Matrix 1 (where repeats are avail-able) as a function of the average final arsenic concentration, Asfave. Thestarred point has a higher than expected standard deviation, and repre-sents the batch test performed at j= 30 mA/cm2 and tCP = 5 min. Thedotted line in a linear fit to non-starred points. . . . . . . . . . . . . . . . 114
3.6 Mixing time kinetics profile for SBGW-2 spiked to Astot = 510 µg/Ltreated at j = 1.1 mA/cm2, q = 150 C/L, tCP = 66 min. Residualarsenic concentration (left axis) and percent arsenic removed (right axis)as a function of mixing time are shown. A dashed horizontal line marksthe WHO limit (10 µg/L). . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.7 Deviation in the final arsenic concentration measured between replicatedbatch tests for Matrix 2 as a function of the average final arsenic concen-tration, Asfave. The deviation plotted between two measurement y1 andy2 is just σ = |y2 − y1|/
√2. The dashed line in a linear fit weighted by
1/As2fave. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.8 Schematic of Three-Electrode Cell used for polarization studies and Elec-tronic Impedance Spectroscopy (EIS). The cell contained a platinum counterelectrode, an Ag/AgCl reference electrode, and an exposed iron disk work-ing electrode (surface area 0.025 cm2). The cell is described more fully inthe text (Section 3.2.9). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.9 A simple Nyquist plot showing complex impedance vector Z(ω). Notethat the Y-axis is negative and that each point on the Nyquist plot is theimpedance at one frequency, ω. In this plot, low frequencies appear onthe right side and high frequencies appear on the left. This is the casewhen the impedance falls as the frequency rises (usually, but not always,true). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.10 A Nyquist plot for a representative electrochemical impedance scan aboutthe open circuit potential in synthetic Bangladesh groundwater (SBGW-1recipe). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.11 Equivalent circuit used in EIS analysis to measure electrolyte resistance ofsynthetic Bangladesh groundwater. Components include the double-layercapacitance, Cdl, the charge-transfer resistance, Rct, and the electrolyteresistance, Re. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.12 Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function of charge density, q, afterECAR treatment at current densities 0.07 - 1.1 mA/cm2 of low phos-phate, zero silicate synthetic Bangladesh groundwater (SBGW-1 recipe -Matrix 1 conditions). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.13 Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function of charge density, q, afterECAR treatment at current densities 0.02 - 100 mA/cm2 of average phos-phate, average silicate synthetic Bangladesh groundwater (SBGW-2 recipe- Matrix 2 conditions). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
viii
3.14 Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function current processing time forMatrix 1 batch experiments for current densities j = 0.07, 0.30, 0.70, and1.1 mA/cm2. Also included are batch tests performed with no externalvoltage (blanks). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
3.15 Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function of charge density, q, for threebatch tests; j = 0.02 mA/cm2 with tCP total = 19.4 hours, j = 1.1 mA/cm2
with the same tCP total = 19.4 hours, and the original Matrix 2 batch testrun for j = 1.1 mA/cm2 using the much shorter tCP total = 1.2 hours. . . . 151
3.16 Arsenic removal (normalized concentration on left vertical axis; percentarsenic removed on right vertical axis) as a function of charge density, q,for two batch tests at j = 0.02 mA/cm2 in which the only difference is adecrease in tCP by 60% for the Fast test relative to the slow tests. Bothbatch tests used Matrix 2 procedures and SBGW-2 water with initialarsenic = 600 µg/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
3.17 Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function of charge density, q, for lowphosphate, zero silicate synthetic Bangladesh groundwater (SBGW-1, us-ing Matrix 1 procedure) and average phosphate, average silicate syntheticBangladesh groundwater (SBGW-2, using Matrix 2 procedure). ECARtreatment occured at j = 1.1 mA/cm2 in both cases. . . . . . . . . . . . . 155
3.18 Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function of contact time for PSA ex-periments using post-synthesis ECAR-generated adsorbents, both freshlyprepared (PSA-Fresh) and aged for 60 minutes (PSA-Aged 60 min). Ad-sorbent was generated using ECAR operating conditions j = 5.0 mA/cm2,q = 175 C/L, tCP = 17.27 min in arsenic-free SBGW-2 water. . . . . . . . 160
3.19 Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function of quiescent contact timefor PSA experiments using post-synthesis ECAR-generated adsorbents,both freshly prepared (PSA-Fresh) and aged for 60 minutes (PSA-Aged60 min), and compared to ECAR treatment at the same current density(j=5.0 mA/cm2) and a similar current processing time (tCP = 16.33 min)as that used to generate the PSA adsorbents. This is a continuation ofFigure 3.18 (note however the change in x-axis units). In all cases, qui-escent contact began after a period of active mixing (tM = 60 min inthe case of ECAR treatment, and tM = 120 minutes in the case of PSAexperiments). Initial arsenic concentration in all three cases was Astot =590 ± 40 µg/L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
ix
3.20 Post-treatment As[III] and As[V] (concentration on left vertical axis; per-cent arsenic removed on right vertical axis) as a function of charge density,q. This batch test was run on SBGW-2 water with operating conditionsj = 1.1 mA/cm2, tCP = 66.58 min, Matrix 2 procedure. These resultswere typical of all current densities in Matrix 2 (i.e. 0.02, 1.1, 5.0, 10, 30,and 100 mA/cm2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
3.21 Comparison of the final As[III] concentration (initial concentration 300µg/L) for SBGW-2 water after ECAR treatment to adsorption with freshlymade post-synthesis ECAR-generated adsorbent (PSA-Fresh) and agedpost-synthesis ECAR-generated adsorbent (PSA-Aged). ECAR treat-ment was performed under similar conditions to the generation of post-synthesis adsorbent. See Section 3.3.8 and Section 3.2.8 for experimentalconditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
3.22 Anodic polarization curve for iron wire electrode in 0.1M KClO4 (thinline) and 0.1M KClO4 + SBGW (thick line) scanned at ν = 0.1 mV/s.Horizontal lines indicate the current densities used in Matrix 1 (dotted)and Matrix 2 (dashed) experiments. Thermodynamic equilibrium poten-tials at 25C for appropriate redox couples are shown for pH =7 (in red) -expected of natural groundwater - and pH = 9 (in blue) - close to the pHof the actual scan. A pourbaix diagram composed by Bang et al. (2005a)assuming solid Fe, Fe(OH)2 and Fe(OH)3 exist was consulted to selectappropriate lines at each pH. If concentration appears in the equilibriumcalculation, a range of concentrations from 10−8M to 1M are shown. . . . 170
3.23 Comparison of the final arsenic concentration for 0.1µm filtered and sed-imentation samples from the Matrix 1 series of experiments. Ts, listedbelow each series of bars, is the quiescent settling time allowed before de-cantation for the settled sample (0.1µm filtered samples were all filteredimmediately after mixing). In the case of series 4 (right), a second samplewas removed and tested from the sedimentation beaker at a later time.The ECAR operating conditions for the above experiments were (1) j =1.1 mA/cm2, q = 23 C/L, tCP = 3 min; (2) j = 1.1 mA/cm2, q = 85C/L, tCP = 5 min; (3) j = 0.07 mA/cm2, q = 25 C/L, tCP = 50 min; and(4) j = 0.70 mA/cm2, q = 15 C/L, tCP = 3 min. Each series has beennormalized to its average initial arsenic concentration to avoid small de-viations due to different starting arsenic levels - the average initial arsenicconcentration was 610 ± 40 µg/L. . . . . . . . . . . . . . . . . . . . . . . 176
3.24 The difference in arsenic concentration between samples filtered througha 0.1µm membrane and samples decanted from the top of a beaker afterECAR treatment (j = 1.1. mA/cm2, q = 100 C/L) following a periodof quiescent settling as a function of the quiescent settling time. Twotests are shown in which the solution was mixed for 30 minutes beforequiescent settling (solid line, solid symbols) and mixed for 60 minutesbefore quiescent settling (dashed line, open symbols). . . . . . . . . . . . . 177
x
4.1 Arsenic concentration measured in filtrate (normalized to the highest av-erage value) after ECAR-treated sample water passed through membranesof different pore size. Errors represent the variability between duplicatetests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
4.2 Scanning electron micrographs of ECAR generated sludge. Image A is ofair-dried sludge collected on a membrane and magnified 500x. The twomarked areas in image A are imaged at 8,000x (image B) highlighting arepresentative square crystal and at 50,000x and 100,000x (image C andD respectively), highlighting a representative area of dry caked powder. . 212
4.3 Energy dispersive spectra taken during SEM imaging of (A) the barepolyvinylidene fluoride membrane (background), (B) a further magnifiedimage of Figure 4.2B (square crystal) centered on a bare section of crystal,and (C) the area shown in Figure 4.2D (caked powder). Au lines appear inall spectra due to the gold coating sputtered onto the sample to facilitateSEM imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
4.4 XRF bicolor map of EGA powder. In this picture, each pixel is coloredred in proportion to the As signal and green in proportion to the Fe signal(pixels with both arsenic and iron appear yellow). This map was measuredfor EGA generated at j = 1.1 mA/cm2. Maps of EGA generated at j =0.02 and 5.0 mA/cm2 were similar. . . . . . . . . . . . . . . . . . . . . . . 215
4.5 Normalized Fe XANES spectra for EGA sludge after ECAR treatment atseveral different current densities (j = 0.02, 0.07, 1.1, and 5.0 mA/cm2)are shown along with Fe XANES spectra for various iron (hydr)oxides ref-erence compounds, including Fe[III](hydr)oxides 2-line ferrihydrite, 6-lineferrihydrite, goethite, lepidocrocite, and akaganeite and Fe(II, III)(hydr)oxides(average oxidation number 2.67) green rust (carbonate) and green rust(sulfate). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
4.6 Iron EXAFS spectra (weighted by k3) for EGA generated at j = 1.1 mA/cm2
compared to reference spectra for iron (hydr)oxides 2-line ferrihydrite (2LFH), goethite, and scorodite. Arrows note significant differences betweenspectra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
4.7 Arsenic EXAFS spectrum (weighted by k3) for arsenic-laden EGA gener-ated at j = 1.1 mA/cm2 compared to reference spectra for As[V] adsorbedonto goethtite, As[V] adsorbed onto 2-line ferrihydrite (2L FH), and themineral scorodite (FeAsO4). Arrows indicate spectral differences. . . . . . 223
4.8 Iron EXAFS spectra (weighted by k3) for EGA generated at current den-sities j = 100 mA/cm2, 5.0 mA/cm2, 1.1 mA/cm2, and 0.02 mA/cm2
(displayed top to bottom). . . . . . . . . . . . . . . . . . . . . . . . . . . . 2244.9 Arsenic EXAFS spectra (weighted by k3) for EGA generated at current
densities j = 100 mA/cm2, 5.0 mA/cm2, 1.1 mA/cm2, and 0.02 mA/cm2
(displayed top to bottom). . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
5.1 Map of Bangladesh highlighting areas visited. . . . . . . . . . . . . . . . . 236
xi
5.2 Percent As[III] relative to initial concentration remaining with time forthree synthetic Bangladesh groundwater batches (WB; open symbols, dot-ted lines) and tube well water sample 12 (TW 12; solid symbol, solid line).The initial As[III] concentration for WB 38 and 45 was not measured, sothe value is assumed to be the average As[III] concentration for all mea-sured water batches, As[III] = 304 ± 41 µg/L. This uncertainty appearsin all subsequent points, since the measured As[III] concentration mustbe divided by the initial As[III] concentration to calculate the percentageAs[III] remaining. The uncertainty in As[III]/As[III]initial for TW 12 isestimated as the percentage decrease in Astot relative to the initial con-centration plus the ICPMS measurement uncertainty for both As[III] andAs[III]initial. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
5.3 Arsenic concentrations before and after ECAR treatment for groundwaterfrom four tube wells in Bangladesh along with SBGW-1 synthetic ground-water (SBGW), plus results from filtration alone (no ECAR) of TW 10after 12 days of settling. Post-treatment concentrations are given for sam-ples filtered immediately after Coulombic dosing (i.e. no mixing, of tM =0 min) and after more than 15 min mixing (tM > 15 min, see Table 5.3 formixing times). Note that only Quick Test arsenic measurements are avail-able for TW 4, tM = 0 min; all other measurements are via ICPMS. Theinitial arsenic concentration for TW 10 and Filtered only TW 10 refersto the arsenic concentration immediately before ECAR treatment and 0.7hrs after collection from the tube well respectively. . . . . . . . . . . . . . 251
5.4 Residual arsenic concentration as function of mixing time (tM ) for SBGW-1 synthetic groundwater (Asinitial = 530 µg/L) and real groundwater fromTube Well 4 (TW 4; Asinitial = 93 µg/L). Note that points overlap at tM= 30 min. All arsenic concentrations were measured using ICPMS exceptfor TW 4, tM= 0 min, which was measured using the Quick Test (hencethe higher error). Error in mixing time is + 3 min for all points due totime of filtration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
xii
List of Tables
1.1 List of countries that have reported arsenic contamination in the ground-water supply as of 2007 (based on (Petrusevski et al., 2007)). . . . . . . . 2
2.1 The iron (hydr)oxides. Adapted from Cornell & Schwertmann (2000) . . . 29
3.1 Synthetic groundwater composition and corresponding mean values andstandard deviations of groundwater in Bangladesh, obtained from the BGSdatabase (values for pH, As[III], As[III]/Astot, HCO –
3 , and Cl – takenfrom special Study Areas; all other values were taken from National Surveydata, using only wells with As > 10 µg/L for data analysis, and settingvalues below detection limits to zero). For each solute, the percentageof wells in the BGS database with a concentration (x) less than that inthe synthetic groundwater 1 and 2 (xs1, xs2) is given. Values for pH, As,As[III], and As[III]/Astot are measured values across water batches, allother ions are spiked values. . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.2 Operating parameters for Matrix 2 tests. . . . . . . . . . . . . . . . . . . . 1183.3 Results of ECAR batch tests in synthetic Bangladesh groundwater (SBGW-
1) containing low phosphate and no silicate. Initial and residual arsenicconcentrations are averaged over replicate batch tests. Boldface indicatesthe lowest residual arsenic concentration achieved for each current density.Errors on arsenic measurements are 10%. . . . . . . . . . . . . . . . . . . 137
3.4 Results of ECAR batch tests in synthetic Bangladesh groundwater SBGW-2, including the average concentrations of phosphate and silicate foundin Bangladesh. Initial and residual arsenic concentrations are averagedover replicate batch tests. Boldface indicates the lowest residual arsenicconcentration acheived for each current density. Errors on arsenic mea-surements are 10%. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
xiii
3.5 Performance metrics to reach the Bangladesh arsenic limit (50 µg/L) andthe WHO arsenic limit (10 µg/L) based on ECAR batch tests in syntheticBangladesh groundwater containing a 1:1 ratio of As[III]:As[V]. Arsenicremoval capacity, iron capacity (assuming all is dissolved as Fe 3+), andtotal treatment time have been calculated based on the WHO limit anddata in Figures 3.12 and 3.13. . . . . . . . . . . . . . . . . . . . . . . . . . 144
3.6 Comparison of arsenic removal capability and arsenic removal capacityusing ECAR treatment versus treatment with freshly made (PSA-fresh)and aged 60 min (PSA-aged) post-synthesis ECAR-generated adsorbents.Arsenic removal capacity has been calculated assuming that all of the ap-plied charge (during ECAR treatment or ECAR-generation of adsorbent)is used to form Fe 3+ ions in solution. . . . . . . . . . . . . . . . . . . . . . 162
4.1 ECAR operating conditions used to generate XAS samples. ECAR op-erating conditions (summarized in Section 1.4.1) include current density,charge density, current processing time (tCP ), and mixing time (tM ). XASsample preparation procedures are described in Sections 4.2.4.2 - 4.2.4.5. . 201
4.2 Chemical analysis of waste EGA sludge or electrode scrapings from ECARtreatment at various operating conditions. Sample preparation is de-scribed in Section 4.2.3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
5.1 Tube well locations, initial arsenic concentration and time of initial aliquot,and list of tests performed on each tube well water sample. . . . . . . . . 239
5.2 Astot, As[III], and As[III]/Astot ratio for TW 12 up to 7 days after collection.2455.3 ECAR operating conditions and results for treatment of tube well samples
and SBGW-1 synthetic groundwater. . . . . . . . . . . . . . . . . . . . . . 2505.4 Groundwater composition of selected tube wells (TW) after 11-19 days
of storage compared to BGS statistics from the same region. Syntheticgroundwater parameters used in batch tests (SBGW-1,2) are added forcomparison (from Table 3.1). BGS statistics include all wells availablein the designated region with Astot ≥ 10 µg/L (standard deviation givenin parenthesis). Italics indicate a concentration that falls outside of therange defined by the BGS regional average value ±the standard deviation. 257
xiv
Acknowledgments
I am indebted to my husband Nathan for his tolerance of endless edits, complaints, and
long days. Thank you Nathan for always believing in me, for your endless support ev-
eryday, and for never letting go.
I am grateful for the constant support of my sister, Kathy Amrose Larsen, who helped
me get to the finish. I am also grateful for a lifetime of support from my parents, Judy
and Fred Amrose, who never gave up on me and allowed me to find my own way. Thanks
also to my twin, Steven Amrose.
I would like to thank my research advisor, Professor Ashok Gadgil, for showing me that
physics can make a difference and for continuing to teach me how. I would like to thank
Professor Bob Jacobsen for never giving up on me and Professor Marjorie Shapiro for her
support of my topic. I am grateful for the support of my committee members, Professor
David Sedlak and Professor Rich Muller, who gave many thoughtful comments. I would
also like to thank Art Rosenfeld, who participated in my qualifying exam and has been
an inspiration to me. A special thanks to Professor Steven Boggs.
I am grateful for the excellent mentoring of Dr. Robert Kostecki, who offered many
hours of advice and patience, and who kindly offered his research space for my thesis
work. I am also grateful for the guidance and advice of Dr. Venkat Srinivasan, Dr. Lara
Gundel, Jonathan Slack and Howdy Goudy. Thanks to postdoctoral researchers Dr.
xv
Laurence Hardwick and Dr. Marek Marcinek for help and patience in the lab. Thanks
to Johanna Mathieu for many useful critiques, comments, advice, and support. I have
had the honor of working with many talented undergraduates on this work, including
Kate Ming, Scott Mclaughlin, Yola Bayram and Emily Desley-Bloom. A special thanks
to Kristin Kowolik. Thanks also to the wonderful student BEAR team - Marc Muller,
Jessica Huang, Debbie Cheng, Marianna Kowalczyk, John Wang, and Michele Itten.
I would like to thank Matthew Marcus, John Bargar, and Sirine Fakra for support and
advice during EXAFS data collection and analysis. Thanks to Dr. Peggy O’Day and
Dr. Andrea Foster for sharing advice and EXAFS spectra. A special thanks to Jasquelin
Pena and Dr. Brandy Toner, who sat down with me to explain EXAFS procedures and
theory and offered wonderful guidance and advice.
My research funding was provided by the LDRD (Laboratory Directed Research and
Development) program at Lawrence Berkeley National Lab, the Blum Center for De-
veloping Economies, the EPA P3: People, Prosperity and the Planet Award, and the
Haas School of Business Sustainable Projects and Solutions Program. I am grateful for
the attention and inspiration provided by George Scharffenberger. I am also grateful
for the administrative assistance of Jane Youn, Lee Borrowman, Donna Sakima, Anne
Takizawa, Claudia Trujillo - I thank you all wholeheartedly for all of your help.
1
Chapter 1
Introduction
1.1 Arsenic contamination
1.1.1 Worldwide
Arsenic in drinking water is a major public health problem threatening the lives
of over 140 million people worldwide (Kahn, 2007; Beck, 2007). The acute toxicity of
arsenic in high concentrations has been known about for centuries, but only recently
has a strong adverse effect on health been associated with long-term exposure to even
very low concentrations (Petrusevski et al., 2007). Over the past few decades, naturally-
occurring arsenic has been discovered in groundwater used to supply drinking water
in many countries on all continents. Well-known high arsenic groundwater areas have
been found in Argentina, Chile, Mexico, China and Hungary, and more recently in West
Bengal (India), Bangladesh and Vietnam (a full list is presented in Table 1.1). Arsenic is
colorless, tasteless, and odorless, and accumulates in the human body over time, making
2
Table 1.1: List of countries that have reported arsenic contamination in the groundwatersupply as of 2007 (based on (Petrusevski et al., 2007)).
Continent Countries Reporting Arsenic in Groundwater
Asia Bangladesh, Cambodia, China (including provinces of Taiwan and InnerMongolia), India, Iran, Japan, Myanmar, Nepal, Pakistan, Thailand, Viet-nam
Americas Argentina, Chile, Dominica, El Salvador, Honduras, Mexico, Nicaragua,Peru, US (including Alaska)
Europe Austria, Croatia, Finland, France, Germany, Greece, Hungary, Italy, Roma-nia, Russia, Serbia, United Kingdom
Africa Ghana, South Africa, Zimbabwe
Pacific Australia, New Zealand
it an insidious threat to public health.
Arsenic poisoning is painful and disfiguring. Skin lesions (known as Keratosis)
often appear first, with a gestation period of 5 to 15 years, commonly causing dark pig-
mentation of the skin and a hardening of the palms and feet (Madajewicz et al., 2008).
Associated symptoms can include weakness, lethargy and fatigue that can prevent work,
chronic respiratory problems, including painful bloody coughing that can be misdiag-
nosed as TB, gastrointestinal problems, cardiovascular disease and neuropathy, which
includes tremors, headaches, vertigo, numbness, and pain (Chowdhury et al., 2000). Ar-
senic may also cause developmental problems in children (Wasserman et al., 2004). The
hardening of the skin associated with skin lesions can be so severe that the flesh may
crack, and gangrene can set in causing the victim to lose his or her limbs (Madajewicz
3
et al., 2008).
Long-term effects of arsenic exposure are known to include liver, lung, kidney
and bladder malfunctions along with hypertension, strokes, and heart disease (Chang
et al., 2004; Chiu et al., 2004; Moore et al., 1997). Recently, exposure to low levels of
arsenic has been linked to a prevalence of Type II diabetes in the US (Navas-Acien et al.,
2008; Kyle & Christiani, 2008). Eventually, exposure leads to cancers of the skin, lungs,
bladder, urinary tract, and kidney. It has been estimated that the risk of dying from
cancer due to chronic exposure to 50 µg/l arsenic is as high as 13 in 1000 (Smith et al.,
2000). The effects of arsenic poisoning are cumulative and can only be reversed up to a
point. In some cases, even complete elimination of exposure cannot reverse the effects1
(Hopenhayn-Rich et al., 1996).
The range of arsenic concentrations found in natural waters is large, ranging
from less than 0.5 µg/L to more than 5000 µg/L(Smedley & Kinniburgh, 2002). Typical
freshwater concentrations are less than 10 µg/L, though occasionally, much higher con-
centrations are found, particularly in groundwater2. Increased knowledge of the cancer
risk led the World Health Organization (WHO) to lower its recommended maximum
limit of arsenic from 50 µg/L to 10 µg/L in drinking water (WHO, 1993).1For more information on the health effects of arsenic exposure, see Saha et al. (1999), Ahsan et al.
(2006), or (Petrusevski et al., 2007) and references therein.2Smedley & Kinniburgh (2002) offers an excellent review of the factors controlling arsenic concentra-
tions in natural waters.
4
1.1.2 Crisis in Bangladesh and West Bengal
The scope and potential ramifications of arsenic contamination are greatest in
Bangladesh, where up to 57 million people daily are exposed to arsenic levels greater
than 10 µg/L, in some cases as high as 2,500 µg/L BGS (2001); Smith et al. (2000).
The bordering region of West Bengal, India is believed to have an additional 6 million
people exposed to arsenic levels of 50 - 3,200 µg/L (BGS, 2001; WHO, 2004). These es-
timates come from a widespread detailed survey of groundwater conducted in 1998/1999
in Bangladesh by the British Geological Survey (BGS), demonstrating that 46% of shal-
low tube wells (< 150 m) have arsenic concentrations exceeding the WHO guideline of
10 µg/L. In 2006, UNICEF similarly reported 32% of wells tested so far exceed 50 µg/L
(UNICEF, 2006). The Water and Sanitation Program South Asia Office (WSP-SA) esti-
mated 20-40 million people in Bangladesh were ingesting unsafe levels of arsenic in 2000
(WSP-SA, 2000). The extent of this problem has led the WHO (Smith et al., 2000) to
describe the situation as ”the largest mass poisoning of a population in history... beyond
the accidents at Bhopal, India, in 1984, and Chernobyl, Ukraine, in 1986.”
The arsenic present in the groundwater of Bangladesh is generally believed to
have been released during weathering and adsorbed onto naturally occurring neoformed
iron oxyhydroxides. Bacteria are thought to have reductively dissolved the iron oxyhy-
droxide coatings using available organic matter, thereby releasing arsenic into solution
(Nickson et al., 2000; BGS, 2001; Harvey et al., 2002; Islam et al., 2004, 2005). Ar-
senic release due to in-situ oxidation of sulfide materials in response to lowering water
levels (due to extended tube well use) has been proposed, but largely dismissed due to
5
the lack of correlation between arsenic and sulfate (Lowers et al., 2007). Contaminated
groundwater has only recently been accessed for drinking. Until the early 1970s, most
of the Bangladeshi population was drinking from shallow hand-dug wells, rivers, and
ponds. These dug wells were not contaminated with arsenic, but were associated with
cholera, diarrhea, and dysentery due to microbial contamination of the surface water
(Kaufmann et al., 2001). In the 1970s and 1980s, UNICEF and other international
agencies responded by helping to install more than 4 million hand-pumped tube wells in
Bangladesh, reportedly converting 80% of the population by 1997 (UNICEF, 1999). In
1983, the first cases of arsenic-induced skin lesions were identified in the West Bengal re-
gion (Saha, 1995) and by 1993, arsenic in tube wells was identified as the cause. In 2006,
the number of people already showing symptoms of arsenic poisoning was estimated at
40,000 with predictions that it could soon rise to 1 million (UNICEF, 2006).
Arsenic poisoning has an even heavier toll in the midst of poverty. Bad nutrition
is known to exacerbate its toxic effects, especially in children (Rahman et al., 2005). In
addition, increased health costs and loss of productivity and income due to symptoms
of arsenic poisoning can become perilous to an individual and their entire family at
subsistence-level living. The welfare costs of exposure to arsenic have been estimated at
US $84 per household per year in neighboring parts of India (Roy, 2008) and are likely
to be similar in Bangladesh. Such a cost can be crippling for a family living on less
than US $1 per day. Other indirect effects include social exclusion because of visible
skin lesions, additional hours of walking to distant potentially safe tube wells (usually
by women and girls), and social strain as neighbors must negotiate for scarce safe water
6
sources that are usually privately owned (WHO, 2000).
1.2 Arsenic remediation and safe water sources
In developed countries, there are several proven technical approaches for remov-
ing arsenic from municipal water supplies. These include: (1) oxidation and stripping,
(2) coagulation, precipitation, and filtration using iron and aluminum salts, (3) lime
softening, (4) ion exchange resins, (5) membrane processes (e.g. reverse osmosis), and
(6) adsorption3 onto commercially-available activated alumina, activated carbon, acti-
vated bauxite, or Granular Ferric Hydroxide (GFH). Each of these has drawbacks for
applicability to highly dispersed, very poor populations with low levels of literacy and
formal education, including high cost and large amounts of toxic waste. Reviews of these
technologies can be found in Bissen & Frimmel (2003) and Ng et al. (2004).
There is no traditional technology in Bangladesh capable of removing arsenic.
Safe water approaches being tried or considered suitable for developing countries fall
into two primary categories - switching to alternative arsenic-free water sources and re-
mediation of arsenic contaminated sources. In Bangladesh, the former category includes
installation of deep tubewells (which are more likely to be arsenic-free), treatment of
surface waters (e.g. Pond Sand Filters or shallow dugwells), and rainwater harvesting.
Arsenic remediation efforts have primarily focused on point-of-use filters for contami-
nated wells. Arsenic removal is often based on adsorption onto a relatively low-cost
chemical sorbent added to the water, such as zero-valent iron, granular ferric hydroxide3Adsorption is defined and discussed more thoroughly in Chapter 2.
7
or activated alumina. Each safe water option varies in capital and recurring costs, levels
of maintenance, water quality, water flow rate, social acceptability and other external
factors.
Of the arsenic-free sources, the installation of deep tubewells is a strong pref-
erence (Hoque et al., 2004). However, deep tubewells are expensive, are only feasible
where deep aquifers exist, and must be installed properly to avoid arsenic contamina-
tion from shallow aquifers. There is also concern that increased pumping from deep
aquifers could draw arsenic in from shallow counterparts (Yu, 2003). Point-of-use filters
have been plagued by high abandonment rates after a short time (Ahmad et al., 2003;
Hoque et al., 2004; Vergara, 2007), due to difficult maintenance or operation, the time
and attention required, and low cultural acceptability. In addition, chemical adsorbents
have limited effectiveness in removing As[III], which makes up about 70-90% of the total
arsenic measured in Bangladeshi tubewells (Bhattacharya et al., 2002).
1.3 A new implementation model for Bangladesh
In light of the repeated problems with household filter adoption and the con-
tinuing prevalence of arsenic in drinking water in Bangladesh, researchers at LBNL have
recognized the need to be innovative in implementation before deciding on technology
design. This interdisciplinary approach has led to the development of a promising new
implementation model for Bangladesh. In this model, an appropriate safe water technol-
ogy is licensed to a community scale clean water center operating under a pubic-private
partnership (PPP). Users pay a small fee to collect treated water from the local clean
8
water center. This fee covers the capital and operating costs of treatment, the cost of ar-
senic monitoring, the capital costs of a small building, the salary of a few local employees,
and a modest profit. A modest profit would provide an incentive for local stakeholders to
efficiently maintain the system, tailor the technology to local conditions, monitor water
quality, and upgrade the technology as appropriate. A community scale is advantageous
because it (1) removes the burden of maintenance from the households, (2) allows for
centralized monitoring of water quality, both ensuring that the treated water continues
to be safe and allowing treatment to be tailored to local water conditions, lowering costs
and waste, and (3) allows for rapid upgrades as new safer or lower cost technology is
developed. Studies have also shown a strong preference for community scale treatment
over household filters (Ahmad et al., 2003).
The PPP is a partnership between a private company, local financial institu-
tions, a local NGO or social marketing company, and a local village government. The
private company builds, operates, and maintains the center until the capital costs are
paid off by the local government. The government can then choose to continue leasing
operation and maintenance to the private company or choose some other provider. Par-
ticipation by the financial institution allows for low upfront costs to the the local village
government as well as startup equity for the private company. The private company
contracts a local NGO or social marketing organization to provide local education and
act as a community liason. Figure 1.1 shows a schematic detailing the roles, cash flow,
and mutual benefit relationships between each of the partnering institutions in the PPP.
The PPP model is an innovative way to balance the tradeoffs between public and private
9
water treatment. Community participation is encouraged through the partnership with
the local government and liaison activities provided by the local NGO. The benefits and
viability of this model have been demonstrated in India by WaterHealth International
(www.waterhealth.com). WaterHealth is a private company providing clean drinking
water to more than a million people in rural Indian villages through publicly-owned,
privately-managed village-scale water treatment centers operating under a PPP.
Private Company
Local NGO
LocalGovernment
FinancialInstitutions
- Provide capital
- Education
- Community liason- Community buy-in
- Ownership of center
or Social MarketingOrganization
Fund under
contract
- Build, operate, and maintain
clean water center
Contract for
operation &
maintenance
Cash flow
Mutual Benefit
Equity for startup
Captal for
installation
Figure 1.1: A schematic detailing the roles, cash flow, and mutual benefit relationshipsbetween each of the partnering institutions in a public-private partnership (PPP). ThePPP is the key to long term viability of a community clean water center. Appropriatesafe water technology would be licensed to the private company.
Safe water technologies must have certain qualities to be compatible with a
community clean water center in rural Bangladesh. Operating costs must be extremely
low for clean water to remain affordable to those making less than US $1 per day.
Using the model of WaterHealth, an ideal retail price would be < US $0.025 per 10
liters, constraining operating costs to be < $0.005 per 10 liters. Waste output should
10
be low, both for environmental protection and due to very stringent and costly waste
disposal requirements for arsenic in Bangladesh (Mathieu et al., 2008). Maintenance
and operation should require skills commensurate with a secondary education or less to
take advantage of the local labor pool. Spare parts and tools for maintenance and repair
should be available locally. Ideally, the system should be scalable, allowing for trialability
in single village units followed by expansion. Last but not least, the technology should
effectively reduce arsenic concentrations in Bangladesh to either the legal Bangladesh
limit (50 µg/L) or ideally, the more conservative maximum recommended limit by the
WHO (10 µg/L).
1.4 ElectroChemical Arsenic Remediation (ECAR)
The implementation model described above allows for a much wider array of
technologies to be viable in Bangladesh than a household filter model. Specifically, an
electricity-based technology could be viable if the treatment costs were low enough to
offset the maintenance, operation, and capital costs of a small scale electricity source
such as Photovoltaics (PV) or battery packs. In light of this possibility, a promising
electricity-based technology known as ElectroChemical Arsenic Remediation (ECAR)
has been designed for rural Bangladesh.
ECAR is a treatment method based heavily on electrocoagulation (EC; dis-
cussed in detail in Section 2.5.2). In EC (and ECAR), a sacrificial iron electrode is
slowly electrochemically dissolved in arsenic-contaminated water. The dissolved iron
quickly hydrolyzes, forming an iron (hydr)oxide sludge with a high affinity for arsenic.
11
In ECAR, the iron (hydr)oxides are allowed further time to complex with arsenic and
coagulate during post-electrolysis mixing (this is not usually the case in EC). Finally,
arsenic-laden iron (hydr)oxides are separated from water, reducing the arsenic concen-
tration to safe levels. For convenience, the iron (hydr)oxide sludge generated during
ECAR is herein referred to as EGA (Electrochemically-Generated Adsorbent).
ECAR technology has a number of potential advantages over chemical adsor-
bents. EGA is manufactured at the time of use, eliminating the need for a costly ad-
sorbent supply chain. In addition, electrochemical processes greatly enhance the arsenic
removal capacity (i.e. arsenic removed per unit iron input) of ECAR relative to the
chemical addition of ferric salts or metallic iron. This is thought to be due to (i) an in-
crease in the rate of rust production (by factors of 10 to 100 over natural rusting rate of
metallic iron), and (ii) the rapid electrochemical oxidation of As[III] to the more favorable
As[V], which binds much more readily to iron-based adsorbents. Thus employing a small
amount of electricity allows for a large increase in efficiency, lower operating costs and
production of far less arsenic-laden waste than most chemical adsorbents. In addition,
the electrodes are self-cleaning if current is alternated, reducing maintenance and elimi-
nating the need to handle strong alkalies and corrosive acids for regeneration (required of
activated alumina and other regenerative adsorbents). Iron, the only consumable input
of ECAR (aside from electricity) is readily available in Bangladesh. Thus ECAR is a
promising potential technology for a community clean water center in Bangladesh.
12
1.4.1 ECAR operating parameters
The parameters controlling ECAR treatment include the current density, j
(mA/cm2), the charge density, q (C/L), and the duration of post-electrolysis mixing,
tM . Other parameters include the duration of filtration, tfilt and the current processing
time, tCP . tfilt is set by the filtration method and tCP is determined by j and q according
to the following (derived in Section 2.5.2):
tCP =q
j(A/V )(1.1)
where tCP is the current processing time, q is the charge density, j is the current density,
and A/V is the ratio of electrode surface area to treatment volume. Each of the control-
ling parameters influence arsenic removal in different, but intertwined ways. These are
briefly laid out below and more fully explained in Section 2.5.2.
Current density, j, controls the interface potential between the iron electrode
and electrolyte (in this case contaminated groundwater). At a given pH, the products
of iron corrosion are highly dependent on the local potential, and thus j influences
the specific iron (hydr)oxide composition of EGA. The specific composition of EGA as a
function of current density is unknown. In addition, the affinity of many iron (hydr)oxide
species for arsenic is unknown. Differences in arsenic affinity could lead to certain regions
of current density that generate more of less efficient EGA.
Charge density, q, describes the net charge added per liter of contaminated
water. Charge can be converted to iron dosage using Faraday’s equation and current ef-
ficiency measurements (Equation 2.19 and 2.21 respectively). Effectively, charge density
13
measures the applied dosage of EGA.
The mixing time, tM , controls the additional duration of contact between ar-
senic and EGA, allowing for further complexation and coagulation with existing EGA.
Adding a separate mixing step allows ECAR to separate the effects of arsenic-EGA con-
tact, which occurs during both dosing and mixing, from the effects of additional EGA
dosage, which occurs during dosing.
1.4.2 Other considerations - treatment time and cost
Each of the ECAR parameters affects important performance considerations,
such as treatment time and operating cost. The total treatment time is ttreat = tCP +
tM + tfilt. tCP is set by q and j according to Equation 1.1, thus the total treatment time
is a function of j and q in addition to tfilt and tM . A household system would require
a short treatment time (< 1-2 hours) to compete with currently available filters. A
community treatment center could effectively store water, extending the viable treatment
time to days. However, given the increased potential for bacterial contamination with
storage time, as well as a desire to prevent unpopular startup delays after maintenance
or unforeseen shut downs, nighttime (< 12 hour) treatment would be advantageous.
Notably, in rural areas, electricity is likely to be intermittent (due to line or photovoltaic
power supply patterns). Thus there is little to gain from decreasing treatment time to
< 5-6 hours, for the water will have to be stored anyway.
Operating costs depend on the amount of energy and iron required to reduce
arsenic in Bangladesh groundwater to safe levels. Iron consumption is directly related
14
to the charge density by Faraday’s law (Equation 2.19). Thus the iron required will be
based on the charge density required to reduce arsenic to safe levels. The charge density
required is based on the affinity of EGA to complex with arsenic and then to coagulate
into particles large enough to be filtered out. Since j and tM affect the composition
and extent of complexation and/or coagulation of EGA, both can indirectly affect the
amount of charge density, and hence iron, required to produce safe arsenic levels. The
energy requirement has similar dependencies. Energy is directly related to the amount of
charge required by ECAR along with the operating voltage4. Thus energy consumption,
as with iron, may be affected by j and tM . The dependancy on operating voltage means
that energy consumption also depends on the resistivity of the contaminated water.
It is clear that each of the controllable ECAR operating parameters (j, q, and
tM ) may affect each of the performance characteristics relevant for viable, effective, and
low-cost arsenic removal in rural Bangladesh.
1.4.3 Previous electrocoagulation research
Previous electrocoagulation research has focused heavily on wastewater treat-
ment. Electricity was first used to treat water in the UK in 1889 (Chen, 2004). In
1909, Electrocoagulation (EC) using aluminum and iron electrodes was patented in the
US as a method for purifying wastewater. Electrolysis sludge treatment plants were in
operation as early as 1911 in California and Oklahoma (Collier, 1912), though all plants
were abandoned thirty years later due to rising costs (Vik et al., 1984). Since then, EC4Note that ECAR runs under constant current conditions, so the operating voltage is not constant
and will tend to drift slightly.
15
has received new interest for different applications due to benefits such as environmental
compatibility, versatility, energy efficiency, safety, selectivity, amenability to automation,
and cost effectiveness (Mollah et al., 2004). EC has been used to treat wastewater con-
taining heavy metals (Pogrebnaya et al., 1995), foodstuff (Chen et al., 2000; Beck et al.,
1974), oil wastes (Biswas & Lazarescu, 1991)[10], textile dyes (Vlyssides et al., 2000;
Do & Chen, 1994; Ibanez et al., 1998; Vlyssides et al., 1999; Gurses et al., 2002; Xiong
et al., 2001), fluorine (Mameri et al., 1998), polymeric wastes (Panizza et al., 2000), or-
ganic matter from landfill leachate (Tsai et al., 1997), suspended particles (Szynkarczuk
et al., 1994; Abuzaid et al., 2002), chemical and mechanical polishing wastes (Belongia
et al., 1999), aqueous suspensions of ultrafine particles (Matteson et al., 1995), nitrate
(Koparal & Ogutveren, 2002; Mishra, 2006), phenolic waste (Phutdhawong et al., 2000),
and refractory organic pollutants including lignin and EDTA (Pouet & Grasmick, 1995;
Chiang et al., 1997; Chen, 2004). Recently, EC has also been proposed to treat potable
water for humus removal, color, and disinfection (Vik et al., 1984; Pouet & Grasmick,
1995; Chen, 2004; Holt et al., 2002).
Proposing EC to reduce arsenic concentrations in water is relatively new. Much
of the focus has been on removing high levels of arsenic (1-1000 mg/L) from wastew-
ater, usually at low pH (Balasubramanian & Madhavan, 2001; Pinisakul et al., 2002;
Mishra, 2006; Hansen et al., 2006; Gomes et al., 2007). Fewer studies have looked at
arsenic removal in potable water at neutral pH(Parga et al., 2005a; Kumar et al., 2004;
Gomes et al., 2007; Arienzo et al., 2002). These studies are discussed in more detail in
Section 2.5.5.
16
The majority of the above studies concentrate on the empirical optimization.
The influence of parameters j and q on removal efficiency in EC have been studied to a
small extent with mixed results. Chen et al. (2000) and Kumar et al. (2004) both reported
that charge density alone (not current density) governs EC efficiency in current density
ranges j = 3 - 30 mA/cm2 and j = 0.65 and 1.53 macmsq respectively, though current
density was found to have a strong influence on the treatment time. Pouet & Grasmick
(1995) reported that current density can effect treatment efficiency, though Chen et al.
(2000) attribute these results to a changing charge density across experiments.
1.4.4 The need for parameter studies in Bangladesh groundwater
Of the experiments to study the influence of j and q, only Kumar et al. (2004)
used arsenic removal at neutral pH as a metric to study EC. The group reported effective
removal of As[III] and As[V], reducing arsenic-spiked Indian tap water from 2 mg/L
to below the WHO limit of 10 µg/L at a charge density of approximately 100 C/L.
However, the water contained no detectable phosphate and low bicarbonate compared to
Bangladesh groundwater (silicate not measured). These ions, all present in abundance
in Bangladesh (BGS, 2001), are known to have a highly detrimental effect on arsenic
removal using iron (hydr)oxides (described in detail in Section 2.4.6). Prior to this work,
no researcher to the authors knowledge has studied the influence of ECAR (or EC)
operating parameters on arsenic removal in the specific groundwater environment of
Bangladesh. Such studies are needed to determine both (1) if ECAR can operate viably,
effectively, and cheaply in the groundwater of Bangladesh and (2) how the parameters
17
might be used to optimize ECAR in the field. To answer each of these questions, it is
necessary to perform both empirical studies (i.e. study the arsenic removal response at
various parameter combinations) and exploratory studies to help uncover, and therefore
control, the underlying arsenic removal mechanism in ECAR.
1.5 Thesis objectives and dissertation structure
Electrochemical arsenic remediation (ECAR) is a promising technology that
is compatible with a community clean water center implementation model and has the
potential to operate at extremely low cost. This thesis presents research into the viability
and effectiveness of ECAR as a low-cost treatment technology for rural Bangladesh.
As an overall theme, this research seeks to demonstrate arsenic removal ef-
fectiveness of ECAR in the groundwater environment of Bangladesh and explore the
influence of several ECAR operating parameters on arsenic removal capacity (i.e. ar-
senic removed per coulomb) acting as a predictor of operating costs. Understanding the
mechanism of arsenic removal as a function of current density is highlighted as potential
means of further optimization. The cause of current density trends is explored using
in-situ techniques (transient voltammetry and EIS) and analysis of reaction products
(via chemical analysis, scanning electron microscopy (SEM), X-ray diffraction (XRD),
and X-ray absorption spectroscopy (XAS)).
Chapters 2, 3, 4, and 5 of this thesis are summarized below. The motivating
ideas for each chapter are given along with a list of specific research objectives (for
Chapters 3, 4, and 5). Chapter 6 provides a summary.
18
1.5.1 Chapter 2: Relevant Scientific Background
This chapter contains a review of key scientific concepts and literature used to
inform the scientific analysis in chapters to come. Basic properties of arsenic and iron
(hydr)oxides are given, followed by a discussion of sorption and adsorption and a review of
arsenic adsorption onto iron (hydr)oxides in the literature. Some basic electrochemistry
is reviewed, followed by a detailed description of the theory of electrocoagulation and a
review of studies using electrocoagulation to remove arsenic from water.
1.5.2 Chapter 3: Chemical and Physical Analysis of Arsenic Complex-
ation with Iron in ECAR
The parameters in ECAR - current density, j, charge density, q, current process-
ing time, tCP , mixing time, tM , and filtration time, tfilt - may have significant and poten-
tially competing influences on performance considerations relevant to rural Bangladesh
- arsenic removal effectiveness, treatment time, and operating costs. To date, only one
study has looked at the influence of operating parameters in electrocoagulation on ar-
senic removal efficiency at neutral pH (Kumar et al., 2004). These experiments were
performed in spiked tap water that differs significantly in composition from groundwater
in Bangladesh. In addition, only a small range of current density was explored (0.65 -
1.53 mA/cm2), ignoring more extreme regions that could produce higher arsenic removal
efficiency. No parameters other than j and q were varied for iron electrodes.
To better understand the influences of ECAR parameters on performance crite-
ria in rural Bangladesh, empirical studies must be performed in the specific groundwater
19
environment of Bangladesh. Exploratory studies to understand the cause of parameter
influence are also essential to better predict and optimize ECAR performance, specifically
with regard to lowering costs (a key consideration for rural Bangladesh).
Chapter 3 Research Objectives
1. Design a laboratory method for ECAR keeping in mind eventual scale-up.
2. Demonstrate the capability of ECAR to reduce high levels of arsenic (>98% sur-
veyed wells) to below the WHO recommended maximum limit of 10 µg/L in syn-
thetic Bangladesh groundwater constituted to mimic key properties relevant to
arsenic removal.
3. Characterize arsenic removal capacity of ECAR with respect to current density, j,
and charge density, q, in synthetic Bangladesh groundwater.
4. Explore the cause of current density effects of arsenic removal capacity and the
mechanism of As[III] removal.
5. Identify trends and tradeoffs in removal capacity and treatment time between op-
erating parameters.
6. Identify an optimal range of current density and charge density for Bangladesh
taking into account arsenic removal effectiveness, treatment time, and minimization
of operating costs.
7. Explore low cost alternatives to membrane filtration in ECAR.
20
1.5.3 Chapter 4: Characterization of Reaction Products
The ECAR process produces arsenic-laden iron (hydr)oxides sludge (i.e. EGA)
as a waste byproduct. The amount and chemical composition of the waste are key to
understanding safe disposal mechanisms and potential disposal costs in Bangladesh. The
sludge can also be analyzed to look for changes in the chemical structure of the EGA
produced at different current densities as well as the chemical structure of arsenic sorbed
to EGA X-ray absorption spectroscopy. This can be used to compare the arsenic removal
mechanism at different current densities and understand the cause of trends in arsenic
removal capacity with current density.
The particle size of EGA sludge was also explored using scanning electron
microscopy (SEM) and membrane pore size experiments. The particle size is a critical
parameter for determining an appropriate low-cost filtration or sedimentation method
for Bangladesh.
Chapter 4 Research Objectives
1. Measure the amount and Fe/As ratio of EGA sludge produced in ECAR at various
current densities.
2. Estimate the EGA particle size using SEM and break through with membranes of
different pore size.
3. Compare the iron (hydr)oxides structure of EGA produced at different current
densities (EXAFS).
21
4. Compare the arsenic-iron (hydr)oxides structure in arsenic-laden EGA at different
current densities (EXAFS).
5. Compare the iron (hydr)oxides structure of EGA to known reference iron oxyhy-
droxide and iron hydroxide compounds (EXAFS).
6. Compare the arsenic-iron (hydr)oxides structure in arsenic-laden EGA to the known
structure of arsenic adsorbed to various iron (hydr)oxides and scorodite (EXAFS).
1.5.4 Chapter 5: ECAR Performance in Real Bangladesh Groundwa-
ter
Neither ECAR nor EC has ever been tested in real Bangladesh groundwater.
Actual groundwater contains a wide range of co-occuring solutes in addition to arsenic
that have been known to affect arsenic removal using iron (hydr)oxides. Thus, it is essen-
tial to demonstrate the ability of ECAR to remove arsenic from real arsenic-contaminated
groundwater from tube well sources in Bangladesh. Since a large sample size is not feasi-
ble in the scope of this thesis, some analysis of co-occuring solutes in real groundwater is
necessary to characterize the representativeness of a small set of samples in the context
of all contaminated Bangladesh groundwater.
In addition, synthetic Bangladesh groundwater spiked with both As[III] and
As[V] showed much higher than expected oxidation rates of As[III] to As[V] during stor-
age. If this effect is present in real groundwater, then storage might be an effective way
to convert high As[III] concentrations to As[V]. Although As[III] oxidation is expected
to be a byproduct of electrochemical processes in the cell, it is not known if this process
22
is thermodynamically or kinetically favorable at all ECAR operating conditions, thus
backup options are desirable. The opportunity to sample real Bangladesh groundwater
was leveraged to test for rapid oxidation of As[III] during groundwater storage.
Chapter 5 Research Objectives
1. Demonstrate the ability of ECAR treatment to reduce arsenic levels from 100 -
500 µg/L in real Bangladesh groundwater to below either the Bangladesh legal
limit (50 µg/L) or ideally, the WHO recommended maximum limit (10 µg/L).
2. Characterize the co-occurring solute composition of tube well water samples and
compare to regional and national survey data from BGS (2001).
3. Measure the level of As[III] to As[V] oxidation in tube well water samples up to
seven days after collection.
23
Chapter 2
Relevant Scientific Background
2.1 Introduction
This chapter contains a review of key scientific concepts and literature used to
inform the scientific analysis in chapters to come. Basic properties of arsenic and iron
(hydr)oxides are given, followed by a discussion of sorption and adsorption and a review of
arsenic adsorption onto iron (hydr)oxides in the literature. Some basic electrochemistry
is reviewed, followed by a detailed description of the theory of electrocoagulation and a
review of studies using electrocoagulation to remove arsenic from water.
2.2 Arsenic in Natural Waters
Arsenic forms a number of inorganic compounds, including hydrides (e.g. ar-
sine), halides, oxides, acids, and sulfides. Of greater than 320 known arsenic minerals,
fewer than 10 are commonly identified in the environment (Welch et al., 1988) - the
24
most commonly reported (excepting scorodite) are arsenic oxides and arsenic sulfides.
Arsenic is very sensitive to mobilization in the typical pH range of groundwaters (pH
6.5-8.5) under both oxidizing and reducing conditions, making it unique among heavy
metalloids and oxyanion-forming elements (e.g. As, Se, Sb, Mo, V, Cr, U, Re)(Smedley
& Kinniburgh, 2002). Arsenic can occur in the environment in the -3, 0, +3, and +5
oxidation state, though in natural waters it is mostly found as oxyanions of trivalent
arsenite [As[III]] or pentavalent arsenate [As[V]]. The main difference between the two
is that arsenate has an oxygen double bond in the molecular structure. In oxygenated
waters, arsenic acid species (H3AsO4, H2AsO –4 , HAsO 2 –
4 , and AsO 3 –4 ) are stable, with
H2AsO –4 dominating at low pH (pH <6.9) and HAsO 2 –
4 at higher pH (H3AsO4 and
AsO 3 –4 occur in extremely acidic and alkaline conditions respectively). In mild reduc-
ing conditions, arsenious acid species (H3AsO3, H2AsO –3 , and HAsO 2 –
3 ) become stable,
with the uncharged species H3AsO3 dominating at a pH less than about pH 9.2. Organic
arsenic forms may be produced by biological activity, but this mostly occurs in surface
waters and very rarely in areas significantly impacted by industrial pollution (Smedley
& Kinniburgh, 2002).
In groundwaters, the ratio of As[III] to As[V] can vary greatly depending on
variations in the abundance of redox-active solids, especially organic carbon, the activity
of microorganisms, and the extent of convection and diffusion of O2 from the atmosphere
(Smedley & Kinniburgh, 2002). Equilibrium calculations predict that As[V] concentra-
tions should be greater than As[III] concentrations in all but strongly reducing conditions
- however this is not always found to be the case in natural waters. In practice, redox
25
equilibrium is achieved slowly and the redox potential tends to be controlled by the
major elements (O, C, N, S, and Fe) (Smedley & Kinniburgh, 2002). Thus in many
areas, including the groundwater of Bangladesh, As[III] has been found under oxidizing
conditions (BGS, 2001; Abdullah et al., 1995). As[III] is more toxic to humans (Korte
& Fernando, 1991; Balasubramanian & Madhavan, 2001; Yamauchi & Fowler, 1994) and
tends to be more difficult to remove from water (see section 2.4).
Most toxic trace metals exist in solution as cations (e.g. Pb 2+, Cu 2+, Ni 2+,
Cd 2+, Co 2+, Zn 2+). These generally become increasingly insoluble as the pH increases
from acidic values, becoming less mobile at the near-neutral pH typical of most ground-
waters. In contrast, most oxyanions, including arsenate, tend to become more mobile
as the pH increases from acidic values because mobility is often determined by sorption
strength rather than solubility (Dzombak & Morel, 1990). This allows anions to persist
in solution in relatively high concentrations even at near-neutral pH. For this reason,
the oxyanion forming elements such as Cr, As, U and Se are some of the most common
trace contaminants in groundwaters (Smedley & Kinniburgh, 2002).
2.3 Iron (Hydr)oxides
Iron (hydr)oxides are ubiquitous compounds in nature and easy to synthesize
through industrial processes and in the laboratory. There are 16 iron (hydr)oxides (Table
2.1), comprised of iron oxides, iron hydroxides, and iron oxyhydroxides (also known as
oxide-hydroxides) - collectively referred to in this work as iron (hydr)oxides. In addi-
tion, there are numerous non-stoichiometric, or mixed, states. All iron (hydr)oxides are
26
composed of Fe together with O and/or OH, with Fe in the trivalent (most common) or
divalent state. The various (hydr)oxides differ in arrangement of basic structural units -
Fe(O,OH)6 or Fe(O)4 - in space. These basic polyhedra can be linked by faces, edges, or
corners. The combination of linkages form the basis of the iron (hydr)oxides mineralog-
ical structure, and result in different characteristic Fe-Fe interatomic differences (Toner
et al., 2008). In some cases, small amounts of anions such as Cl – , SO 2 –4 , or CO 2 –
3 may
also participate in the structure.
The oxyhydroxides can be dehydroxylated to their oxide counterparts. This is
due in part to the similarity between anion frameworks - often only the rearrangement
of cations and the loss of OH are required to effect a transformation (Cornell & Schwert-
mann, 2000). The Fe[III] (hydr)oxides exhibit low solubility (leading to high stability)
and brilliant colors, often in orange to red shades. Due to their high energy of crystal-
lization, iron (hydr)oxides often form minute crystals with a high specific surface area,
often > 100 m2/g (Cornell & Schwertmann, 2000), making them effective sorbents for a
large range of dissolved ions, molecules, and gases.
Iron oxide minerals tend to form in poorly crystalline states with high specific
surface area such as amorphous-Fe(OH)3 but, over time, undergo transformations to
more crystalline forms such as goethite or hematite (Dixit & Hering, 2003; Benjamin,
2002). The time scale of the transformation can depend on temperature, pH, and the
presence of co-occuring solutes (Cornell & Schwertmann, 2000; Ford, 2002).
The following is a brief description of the compounds listed in Table 2.1, based
heavily on Cornell & Schwertmann (2000). More detailed descriptions (along with an
27
excellent treatment of many iron (hydr)oxides properties) can be found in Cornell &
Schwertmann (2000).
Goethite, α-FeOOH, occurs in rocks and is one of the most thermodynamically stable
iron (hydr)oxides at ambient temperatures, making it either the first (hydr)oxide
to form, or the end member of many transformations. As a powder, Goethite is
yellow. Its structure is based on hexagonal close packing of anions.
Lepidocrocite, γ-FeOOH, is orange colored and occurs in rocks, soils, biota, and rust
and is often an oxidation product of Fe 2+. Its structure is based on cubic close
packing of anions.
Akaganeite, β-FeOOH, occurs rarely in nature and has a structure based on body
centered cubic packing of anions, often containing a low level of either chloride or
fluoride. It appears brown to bright yellow.
Schwertmannite, Fe16O16(OH)y(SO4)z · nH2O, has the same basic structure as aka-
ganeite, but contains sulphate instead of chloride ions. It frequently occurs in
nature as an oxidation product of pyrite, but is unlikely to play a role in this work
due to the relatively low concentrations of sulphate in the environment of interest.
δ-FeOOH (synthetic) and its poorly crystalline mineral form Feroxyhyte, γ′-FeOOH,
are reddish-brown and ferrimagnetic with hexagonal close packing of anions. Fer-
oxyhyte is very rare, but found in certain surface environments.
Ferrihydrite, Fe5HO8 · 4H2O, is reddish-brown and widespread across many surface
environments. Ferrihydrite is often termed ”amorphous iron oxide” or ”hydrous
28
ferric oxide (HFO).” Unlike other iron (hydr)oxides, it exists exclusively as nano-
crystals and, unless stabilized in some way, transforms with time into more stable
and more crystalline iron (hydr)oxides. It is an important precursor, and thus very
important for this work. Structurally, it consists of hexagonal closely packed anions
and is a mixture of defect-free and defective structural units. The composition,
especially with respect to OH and H2O, seems to be variable - thus the formula
Fe5HO8 · 4H2O is often used as a preliminary starting point.
Bernalite, Fe(OH)3, is a greenish hydroxide that to date, has been found only in two
locations (Birch et al., 1993; Kolitsch, 1998).
Fe(OH)2 does not exist as a mineral, but is readily oxidized, developing into greenish
blue green rust, or black magnetite.
Hematite, α-Fe2O3, is the oldest known iron (hydr)oxides mineral and is found abun-
dantly in rocks and soils. It is blood-red if finely divided, or black to grey if coarsely
crystalline. Its structure is based on hexagonal close packing of anions, and, like
Goethite, it is extremely stable and often the end member of transformations of
other iron oxides. Hematite is also known as ironIIIoxide and ferric oxide.
Magnetite, Fe3O4, is a black, ferrimagnetic mineral containing both Fe[II] and Fe[III]
with an inverse spinel structure. It is the principal component in loadstones, the
earliest known permanent magnets.
Maghemite, γ-Fe2O3, is a red-brown, ferrimagnetic mineral with nearly the same struc-
ture as magnetite, but containing cation deficient sites. It occurs in soils as a
29
Table 2.1: The iron (hydr)oxides. Adapted from Cornell & Schwertmann (2000)
Oxyhydroxides and Hydroxides Oxides
Goethite α-FeOOH Hematite α-Fe2O3
Lepidocrocite γ-FeOOH Magnetite Fe3O4(Fe IIFe III2 O4)
Akaganeite β-FeOOH Maghemite γ-Fe2O3Schwertmannite Fe16O16(OH)y(SO4) z · nH2O β-Fe2O3δ-FeOOH ε-Fe2O3Feroxyhyte γ′-FeOOH Wustite FeOHigh Pressure FeOOHFerrihydrite Fe5HO8 · 4H2OBernalite Fe(OH)3Fe(OH)2Green Rusts Fe III
x Fe IIy (OH)3x+2y-z(A – )z;A
–−−Cl –; ( 1
2) SO 2 –
4
weathering product of magnetite or a product of heating other iron (hydr)oxides.
β-Fe2O3, ε-Fe2O3, and High Pressure FeOOH are rare laboratory compounds obtained
only in high vacuum, high temperature (170 - 750 C), and/or high pressure (8
Gpa).
Wustite, FeO, is black, contains only divalent Fe, and is usually non-stoichiometric (O-
deficient). The structure is similar to NaCl and is based on cubic close packing of
anions.
Green Rusts consist of layers of Fe IIOH octahedra in which some Fe[II] is replaced by
F(III) and, to maintain neutrality, anions suh as Cl – and SO 2 –4 are bound between
the layers. Green rusts commonly form as corrosion products.
30
2.4 Arsenic sorption onto iron (hydr)oxides
2.4.1 Theory
Adsorption is the accumulation of a substance at or near an interface relative to
its concentration in the bulk solution (Benjamin, 2002), also called surface complexation.
The substance that adsorbs is called the adsorbate, and the solid to which it binds
is called the adsorbent. Although adsorption is technically distinct from absorption,
which is the accumulation of substance in the interior of a non-aqueous phase, the two
processes are often referred to together using the terms sorption, sorbate, and sorbent.
Sorption can also include surface precipitation (Sposito, 1986), which can occur when
the accumulation of adsorbate at the surface is sufficient for a new iron-arsenic phase to
precipitate on the surface of the iron (hydr)oxide (i.e. a regular stoichiometric crystal-
like structure that can grow as opposed to a single surface layer of adsorbed ions).
Distinct from all of the above is the formation of a solid-solution phase, such as ferric
arsenate (FeAsO4), in which arsenic is incorporated into the bulk phase rather than
just the surface. For this work, sorption will refer to a combination of adsorption,
absorption and/or surface precipitation. The term adsorption will be reserved for surface
complexation only, distinct from surface precipitation. Solid-solution formation will be
referred to independently.
The mechanism of adsorption is often chemical interactions between the ad-
sorbate and surface rather than electrostatic interactions. When this is the case, the
adsorbate is said to be chemically or specifically adsorbed (often called chemisorbed). If
31
the adsorbate is attracted to the surface solely through electrostatic interactions, then it
is said to be physically or non-specifically adsorbed. The interaction between dissolved
arsenic oxyanions and iron (hydr)oxides occurs at the iron (hydr)oxide surface. In the
presence of water, the surface of an oxide is assumed to be hydrated and covered with a
layer of OHx groups (either from the oxide structure or bound water molecules). Func-
tional hydroxyl groups on a surface contain the same donor atoms as functional groups
in soluble ligands, as so behave in the same way. The groups contain ionizable atoms
and so can protonate and deprotonate as a function of pH (i.e. acid/base reactions).
In the case of arsenic oxyanions, the main mechanism of adsorption is ligand exchange
(Stollenwerk, 2003); the surface hydroxyl group is exchanged for another ligand. Ligand
exchange can occur at one or two surface hydroxyl groups on the oxide, designated by
S−OH, according to the following:
S−OH + L −−−− S−L + + OH− (2.1)
2 S−OH + L −−−− S2L 2+ + 2 OH− (2.2)
These bonds are strong and are known as inner-sphere complexes. If the bond occurs
at one hydroxyl group (Equation 2.1), it is known a monodentate complex. If it occurs
at two hydroxyl groups (Equation 2.2), it is a bidentate complex. Since adsorption is
coupled with the release of OH- in both cases, complexation tends to be favored by lower
pH values. A complex in which the adsorbate retains its waters of hydration (while still
engaging in a chemical bond) is known as an outer-sphere complex.
The capacity of a suspended solid surface adsorbent to adsorb a species of ion
from solution depends on many factors, including the number of available sites on its
32
surface, the affinity of adsorbate for each site, as well as on the amount of adsorbate
available near the surface. The availability of sites on the adsorbent depends on the
adsorbent surface area, the protonation state of the surface hydroxyl groups (which
depends on pH), and the concentration of competing ions for the same sites. The affinity
of the adsorbate for a site depends on local electrostatic conditions (partially created by
the protonation state of the surface hydroxyl groups) as well as the thermodynamic
favorability of a chemical reaction. Each of these factors can limit and/or control the
quantity of adsorbate removed from the solution through adsorption.
The amount of material adsorbed per unit amount of adsorbent is called the
adsorption density. Any model that relates the amount of adsorbate at the surface to that
in the solution in equilibrium systems at constant temperature is called an adsorption
isotherm. Adsorption of As[III] and As[V] to iron oxides and iron hydroxides can be
described by a Langmuir isotherm (Ferguson & Gavis, 1972; Anderson et al., 1976;
Gupta & Chen, 1978; Pierce & Moore, 1982; Sun & Doner, 1996). The form of the
Langmuir isotherm for a simple system with a single adsorbate is:
qA =KadsA
1 +KadsAqmax (2.3)
qA = adsorption density (moles adsorbed per gram of adsorbent)
Kads = constant defining affinity of adsorbent for adsorbate
A = activity × aqueous concentration of adsorbate A
qmax = maximum adsorption density.
When KadsA << 1, the isotherm becomes linear: qA ≈ KadsAqmax. Over this
limited range of adsorbate concentrations, one can expect the percentage of adsorbate
33
removed from solution to be approximately constant. As KadsA >> 1, qA ≈ qmax,
and the percentage of adsorbate removed will decrease will increasing adsorbate concen-
tration.
If multiple adsorbates are in solution, they may compete for adsorption sites
on a single adsorbent. In the case of j adsorbates, the Langmuir adsorption isotherm for
any species i becomes:
qi =Kads,iai
1 +∑
jKads,jajqmax (2.4)
Kads,i = constant defining affinity of adsorbent for adsorbate i
ai = A for adsorbate i
See Benjamin (2002), chapter 10, or Stumm & Morgan (1996), chapter 9, for more details
on the derivation and behavior of isotherms.
The surface of the adsorbent can acquire electric charge though chemical re-
actions via ionizable functional groups (i.e. acid/base reactions, dependant on pH),
isomorphic replacements of atoms in the surface lattice, and adsorption. The charge
can affect the affinity of the adsorbent for the adsorbate. To take this into account, the
constant Kads can be modified by an exponential term that is a function of the surface
potential (Benjamin, 2002; Stumm & Morgan, 1996). In general, a more charged surface
will have an increased affinity for opposite-charge adsorbates and a decreased affinity
for same-charge adsorbates. Recall from Section 2.2 that the stable forms of arsenate
between pH 6-9 are negatively charged (H2AsO 2 –4 and HAsO 2 –
4 ) while the stable form
of arsenite is neutral (H3AsO 03 ).
34
If the adsorption density becomes high enough, it is possible for the adsorbate
to form a surface precipitate. In theory, this would eliminate the limiting capacity of the
solid for adsorbate (i.e. adherence to Equations 2.3 and 2.4), since ions from the bulk
could always (subject to solubility equilibrium) be incorporated into a new precipitate
layer as opposed to competing for adsorption sites. This appears to provide a method for
distinguishing adsorption from surface precipitation by measuring the isotherm behavior
(a surface precipitate should have no qmax). However, in practice it almost impossible
to distinguish the two states from isotherms alone. Surface precipitation is likely to be
accompanied by adsorption to other parts of the surface with the sum of both behaviors
appearing in the isotherm. In addition, if the affinity of the adsorbent for the adsorbate
varies across the surface (as it often does), the isotherm may appear to increase indefi-
nitely as strong binding sites become saturated but weaker binding sites continue to come
into play at higher adsorbate concentrations - the isotherms become nearly impossible to
distinguish from the expected isotherm of surface precipitation. Sensitive spectroscopic
techniques, such as X-ray Absorption Spectroscopy (XAS), are often used to clarify the
ambiguity instead.
Sorption of arsenic in a lab setting is commonly achieved by either coprecipita-
tion or post-synthesis adsorption (PSA). In coprecipitation, iron (hydr)oxides are formed
in the presence of arsenic oxyanions. In PSA, iron (hydr)oxides are formed prior to the
addition of arsenic. The adsorbent aging period before the arsenic is added allows some
time for coagulation of the adsorbent and crystallite growth to occur, both processes
which tend to decrease the number of adsorption sites (Waychunas et al., 1993). If,
35
however, a sorbing ion is present in solution during the hydrolysis and precipitation of
iron (hydr)oxides (as in coprecipitation), then the ion may sorb to surface sites before
aggregates are formed. This maximizes the number of available surface sites, increasing
the sorptive capacity of the iron (hydr)oxides (Fuller et al., 1993). For modeling of natu-
ral systems, coprecipitation is often more appropriate, since natural iron(III)hydroxides
are often formed from oxidation of Fe 2+ ions in the presence of arsenic (Roberts et al.,
2004). However, numerous arsenic remediation technologies exist that are based on either
method, and therefore both coprecipitation and PSA are commonly used in experimen-
tation.
2.4.2 Review of arsenic adsorption literature
A number of iron (hydr)oxides adsorbents have been effectively used to remove
arsenic compounds from water. These include pre-formed iron (hydr)oxides (Pierce &
Moore, 1982; Goldberg, 1986; Waychunas et al., 1993; Fuller et al., 1993; Fendorf et al.,
1997; Grossl et al., 1997; Raven et al., 1998; Manning et al., 1998; Jain et al., 1999)
and coprecipitation with ferric and ferrous salts (Hering et al., 1996; Roberts et al.,
2004; Sorg & Logsdon, 1978). Zero-valent iron (ZVI) filings have also been effective at
removing arsenic compounds from water (Su & Puls, 2001; Lackovic et al., 2000; Bang
et al., 2005b,a), presumably due to adsorption onto iron corrosion products. In direct
comparisons, coprecipitation has been found to be more efficient than arsenic adsorption
onto pre-formed iron (hydr)oxides (Fuller et al., 1993; Edwards, 1994), referred to here
as PSA. However, in addition to being more pronounced at high aorbate-sorbent ratios,
36
these effects were found to be slowly reversible (Hering et al., 1996), with the excess As[V]
gradually releasing back into solution until the adsorption density of coprecipitation
converged (within about 20 days) with that of adsorption alone (Fuller et al., 1993).
Thus it is more appropriate to say that coprecipitation is initially more efficient than
PSA. In general, higher removal efficiencies can be achieved with increased coagulant or
adsorbent dosages for both coprecipitation and PSA (Hering et al., 1996; Cheng et al.,
1994; Edwards, 1994; Gulledge & Oconnor, 1973).
Arsenate and arsenite adsorption have been found to obey a Langmuir isotherm
on iron hydroxide and aluminum hydroxide for initial arsenic concentrations in the range
10−5 − 10−3M at pH 6.0-7.5 (Anderson et al., 1976; Dixit & Hering, 2003; Ferguson &
Anderson, 1974). Gupta & Chen (1978) reported similar results for adsorption on alu-
mina and bauxite for initial pH 6.5-8.5 and initial arsenic concentrations of 5 - 130 µM.
However, Holm et al. (1979) found adsorption of arsenite on river sediments was linearly
dependent on concentration while arsenate adsorption followed a Langmuir isotherm. Ini-
tial arsenic levels were 10−5−10−4M and pH was not mentioned. Pierce & Moore (1982)
found similar linear behavior for high initial concentrations of 33.4-667 µM for both arse-
nate and arsenite adsorption on amorphous-Fe(OH)3, though a Langmuir isotherm was
obeyed at concentrations 0.667-13.3 µM in every case. All agree that Langmuir is obeyed
for arsenic concentrations found in Bangladesh, where 98% of wells contain Astot < 8
µM (BGS, 2001).
The typical adsorption density for As[V] on pre-formed hydrous ferric oxide
ranges from 0.075 to 0.16 mg As[V] per mg Fe in the pH range 4-7 (Driehaus et al., 1998;
37
Wilkie & Hering, 1996). For As[III], the adsorption density varies from 0.06 to 0.07 mg
As[III] per mg Fe at pH 7.0 (Pierce & Moore, 1982; Wilkie & Hering, 1996). Roberts et al.
(2004) studied coprecipitation in synthetic Bangladesh groundwater through chemical
addition of Fe[II] or Fe[III] as FeSO4 and FeCl3. For groundwater free of P and Si, As[V]
removal was similar using either Fe[II] or Fe[III]; the addition of 2-2.5 mg Fe/L was
sufficient to reduce 500 ppb to less than 50 ppb. Less than 1/10 of either Fe[II] or Fe[III]
was required to remove As[V] compared to As[III], in agreement with Meng et al. (2002).
Borho & Wilderer (1996) found similar arsenic removal for Fe[II] and Fe[III] in waters
low in As, P, and Si. Mamtaz & Bache (2001) were able to reduce arsenic in synthetic
groundwater (1.2 mM NaHCO3) with no P, Si, Ca and Mg from 200 ppb As[III] by 47%
and 43% using 4 mg/L Fe[II] and Fe[III] respectively after 1 day of settling.
The strength of arsenic adsorption onto iron (hydr)oxides has been widely stud-
ied using several experimental techniques, mostly pointing towards formation of an inner-
sphere complex. Some studies focus on the effects of Ionic Strength (I) and pH on arsenic
adsorption. The ionic strength quantifies the composite effects of all charged ions in so-
lution, which can impact certain chemical reactions. Changes in I affect the electrostatic
forces near the mineral surface (Davis & Kent, 1990), and thus affect adsorption via
outer-sphere complexes much more strongly than inner-sphere complexes. Hsia et al.
(1994) and Grossl & Sparks (1995) show essentially no change in the concentrations of
As[V] adsorbed by ferrihydrite and goethite for I in the range of 0.005 to 0.1 M, con-
sistent with an inner-sphere complex. Studies of changes in the isoelectric point (IEP)
of oxides (the solution pH at which the net charge on the particle is zero) also indicates
38
an inner-sphere complex for both As[III] and As[V] on goethite, ferrihydrite, gibbsite,
and amorphous Fe(OH)3 (Anderson & Malotky, 1979; Ghosh & Teoh, 1985; Hingston
et al., 1971; Jain et al., 1999; Manning & Goldberg, 1996a). Finally, Extended X-ray
Absorption Fine Structure (EXAFS; discussed more fully in Chapter 4) has confirmed
the formation of binuclear bidentate inner-sphere complexes for As[V] adsorbed onto
ferrihydrite (Waychunas et al., 1993; Manceau, 1995) and goethite (Fendorf et al., 1997).
Manning et al. (1998) used EXAFS to show similar inner-sphere bonding of As[III] on
goethite, confirming results using Fourier Transform IR Spectroscopy (FTIR) by Sun &
Doner (1996). There is some evidence that, in addition to inner-sphere complexes, As[III]
may also form outer-sphere complexes on iron oxides (Goldberg & Johnston, 2001).
Surface precipitation at high concentrations has been suggested as a mecha-
nism for arsenate and phosphate sorption onto ferrihydrite due to slow sorption and
high sorptive capacity (Dzombak & Morel, 1990; van Riemsdijk et al., 1984). However,
EXAFS analysis by Waychunas et al. (1993) revealed no such phases in coprecipitation
or post-synthesis adsorption of As[V] onto Ferrihydrite at pH 8.0 and an ionic strength
of 0.02M during the first seven weeks of aging.
2.4.3 Arsenic removal via zero valent iron
The mechanism of arsenic removal in oxic conditions (DO >0.1 mg/L) via zero-
valent iron (ZVI) is adsorption of As[V] and As[III] oxyanions to ferric oxide precipitates
resulting from corrosion of Fe(0) and subsequent oxidation of Fe 2+ (Bang et al., 2005b,a).
Mono- and bi-dentate complex formation of arsenic with iron oxides produced from iron
39
corrosion has been reported by Lackovic et al. (2000), Farrell et al. (2001) and Su & Puls
(2001). The reactions can be described by the following:
2 Fe 0 + O2 + 4 H + −−−− 2 Fe 2+ + 2 H2O (2.5)
4 Fe 2+ + 4 H + + O2−−−− 4 Fe 3+ + 2 H2O (2.6)
Fe 3+ + 3 H2O −−−− Fe(OH)3 + 3 H + (2.7)
The rate of ferric hydroxide addition to the solution is directly related to the corrosion
rate. Higher dosages of ferric hydroxide lead to more arsenic removal (Bang et al., 2005b;
Mishra & Farrell, 2005). The corrosion rate is strongly dependent on DO concentration
(Equation 2.5) as well as pH, with the required contact time for removal increasing
significantly as DO is decreased or pH is increased (Bang et al., 2005b; Lackovic et al.,
2000). Both effects are strong - after 9 hours of oxic reaction with 1g/L Fe(0) filings,
99.8% removal was achieved for initial As[V] concentrations of 100 µg/L at pH 6, while
only 55.5% and 2% were removed at pH 7 and 8 respectively (Bang et al., 2005b). In
the same time, 99.8% of As[V] was removed at pH 6 in solution open to air compared to
only 9% in pH 6 anoxic solution. The corrosion rate (measured by the total iron content
in solution) increases rapidly for about 10 minutes before approaching steady state. In a
0.04 M NaCl solution under atmospheric conditions at pH 7, Bang et al. (2005a) reported
a steady state corrosion rate of about 0.0875 mg Fe/min/L. It is important to note that
the corrosion rate is very sensitive to surface area.
In the limit of a low ratio of adsorbing species to adsorption sites, the arsenic
removal rate is limited by mass transfer to adsorption sites, and is therefore exhibits first
40
order dependance on arsenic concentration (Farrell et al., 2001; Melitas et al., 2002b).
However, at high solution concentrations, removal becomes limited by the rate of adsorp-
tion site generation (related to the corrosion rate and ferric hydroxide formation rate)
and is therefore zero order in arsenic concentration (Farrell et al., 2001; Melitas et al.,
2002b).
On a weight basis, ZVI is approximately as efficient as coprecipitation with
FeCl3. Over 200 days of operation, Bang et al. (2005b) were able to remove 18,000 mg
of arsenic using 400g Fe(0) filings - a capacity of 0.045 gAs/g Fe (including a 1:3 ratio of
As[III] to As[V]). However, breakthrough of arsenic began to occur after only 100 days,
indicating that filings would have to be replaced long before the capacity was reached.
As[III] removal is generally less than As[V] removal due to a lower adsorption
affinity of Fe(OH)3 for As[III] at pH < 8 (Bang et al., 2005b; Meng et al., 2000). However,
recent research by Hug & Leupin (2003) shows that As[III] can be oxidized in parallel to
the oxidation of Fe[II]. This was attributed to intermediates formed during the reduction
of O2 with Fe[II]. The addition of 90 µM Fe[II] to 6.6 µM As[III] air-saturated solution
resulted in 25 - 30% oxidation of the As[III] at neutral pH.
2.4.4 Effect of pH and speciation on arsenic sorption
As discussed in Section 2.4, arsenic adsorbs to iron (hydr)oxides by ligand
exchange with OH and OH +2 surface functional groups, forming an inner-sphere complex
(Stollenwerk, 2003). This requires an incompletely dissociated acid (i.e. H2AsO –4 ) to
provide a proton for complexation with the surface OH group, forming H2O and providing
41
a space for the anion (Hingston et al., 1971). The energy required to dissociate the weak
acid at the oxide surface varies with pH, and thus the amount of arsenic adsorbed varies
with pH.
The effect of pH on arsenic adsorption differs between As[III] and As[V], re-
flecting the difference in net charge of the thermodynamically stable species with pH.
Typically, metal oxyanions shift from being mostly adsorbed at low pH to mostly dis-
solved at high pH, as pH increases through a critical range 1-2 pH units wide (Benjamin,
1983). Although the location of the critical pH range can be adjusted by changing the
system composition, the general shape of the pH-adsorption edge is similar in almost all
systems (Benjamin, 1983). Arsenate behaves like metal oxyanions, its adsorption tends
to be greatest at low pH and decreases with increasing pH (Pierce & Moore, 1982),
attributable to the attraction between a positively charged surface at low pH and the
negatively charged predominant arsenate species (H2AsO –4 ) between pH 2.2 and 6.9
(Stollenwerk, 2003). At pH values above 6.9, the predominant species becomes more
negative (HAsO 2 –4 ), but the surface becomes less positive, making adsorption less favor-
able. The pH dependence of As[III] is different. The predominant As[III] sprecies at pH
less than 9.2 is the neutral H3AsO 03 , more readily able to donate protons to the surface
OH group than the negatively charged As[V] species.
The relative binding strength of As[III] compared to As[V] depends on the type
of iron oxide, the solution pH, and the solution concentration of each species (Pierce
& Moore, 1982; Raven et al., 1998; Manning et al., 1998; Jain et al., 1999). Arsenic
adsorption has been observed to be more highly dependent on its oxidation state than
42
pH within the range of 5.5-7.5 (Ferguson & Anderson, 1974). As[V] removal during
coprecipitation with alum, ferric chloride (Fe 3+), and ferrous sulfate (Fe 2+) is more
efficient than As[III] removal under comparable conditions (Hering et al., 1996; Gulledge
& Oconnor, 1973; Sorg & Logsdon, 1978; Roberts et al., 2004; Meng et al., 2002; Shen,
1973; Leckie et al., 1980; Meng et al., 2000). However, for PSA experiments on goethite
and ferrihydrite (amorphous-HFO), a crossover pH is often observed with As[V] more
efficiently adsorbed below the crossover pH and As[III] more efficiently adsorbed above
it. The value of the crossover pH and trends with total arsenic concentration vary widely.
Dixit & Hering (2003) report the crossover pH for HFO as 6.5 with 50 or 100 µM total
arsenic and 8.5 with 10 µM total arsenic. Similar observations were made by Raven
et al. (1998) for HFO with 534 and 1600 µM total arsenic, though at 267 µM As[III]
was found to be more efficient from pH 3 -11, inconsistent with the results of Dixit &
Hering (2003). Pierce & Moore (1982) report a crossover at pH 6.5 and 7 for total arsenic
concentrations of 1.33 µM and 13.3 µM respectively, reversing the trend seen by Dixit &
Hering (2003). For goethite, the crossover pH was found to be 6.0 with 100 µM, 7.0 with
50 µM, and 7.0-8.5 for 25 µM by Dixit & Hering (2003). Manning et al. (1998) reports
a crossover pH for goethite of 5.0 at 267 µM and Sun & Doner (1996) report pH 6.0 at
1250 µM total arsenic. These variations may reflect differences in the total sorbate and
sorbent concentrations used in the experiments (Dixit & Hering, 2003). The variation in
measurements, including at lower arsenic concentrations relevant to Bangladesh, suggest
that it is possible for As[III] adsorption to be comparable or even greater than As[V]
adsorption at pH 7, the typical pH of Bangladesh groundwater (see Section 3.2.3 below).
43
As[III] and As[V] can also compete when in solution together. Jain & Loeppert
(2000) found that As[III] had little effect on adsorption of As[V] below pH 6, but caused
a decrease in As[V] adsorption by as much as 25% at pH greater than 6 due to strong
adsorption of As[III]. In contrast, As[V] caused a decrease in adsorption of As[III] by
15-20% between pH 4 and 6; however by pH 9, As[V] decreased adsorption of As[III] by
less than 5%. However, the pH behavior described above is dominant over competitive
adsorption when the concentration of each anion is low relative to the iron precipitate,
below 0.1 mol As/kg HFO (Jain & Loeppert, 2000).
2.4.5 Kinetics of arsenic sorption
The time dependence of ion sorption by mineral surfaces is often characterized
as a two step process (Sposito, 1986; Davis & Kent, 1990) with initial rapid sorption
followed by slower sorption. The slower sorption step can be caused by several different
types of processes, including the formation of a surface precipitate (Davis et al., 1987),
diffusion to adsorption sites in the internal porosity of minerals (Stollenwerk & Kipp,
1990), diffusion into the inter-particle porosity of aggregated particles (Wu & Gschwend,
1988; Willett et al., 1988), or the rate of colloid aggregation and coagulation (Honeyman
& Santschi, 1991). Adsorption onto pre-formed iron (hydr)oxides (PSA) tends to take
longer because, as aggregates form, adsorption sites on the surfaces of crystallites become
buried within large clusters of particles (Willett et al., 1988). Adsorbing ions react first
with the exposed exterior surface sites, but must diffuse within the aggregates before
a steady-state concentration of the adsorbing ion can be obtained (Fuller et al., 1993).
44
The rate of adsorption tends to be higher during coprecipitation because diffusion into
aggregates or clusters is no longer necessary. This is enhanced by the presence of adsorbed
ions, which can slow the rates of coagulation and crystallite growth (O’Melia, 1987;
Anderson et al., 1985) allowing high rates of adsorption to continue longer. However,
as the coprecipitate ages, eventual crystallite growth and increased corner-sharing by
Fe octahedra cause desorption of ions (those that cannot be incorporated in a stable
three-dimensional structure) (Fuller et al., 1993). This means that some arsenic could
be expelled from the crystallizing matrix as it ages, slowing or even reversing the net
rate of arsenic adsorption.
The kinetics of adsorption onto amorphous ferrihydrite, and many other oxides,
also includes an initial rapid step, typically between 5-30 min, and is sometimes followed
by a slower step during which the adsorption density may increase at a rate of a few
percent per day (Benjamin, 1983; Kurbatov & Wood, 1952; Posselt et al., 1968). Fuller
et al. (1993) reported an initial rapid uptake of As[V] (<5 min) by ferrihydrite in solution
at pH 8 followed by a slow uptake (over 8 days) consistent with diffusion to adsorption
sites within aggregates or colloidal particles (pH was maintained throughout the exper-
iment). (Grossl et al., 1997) used a pressure/jump relaxation technique on goethite to
find that the formation of monodentate ferric-arsenate surface complexes form on a time
scale of milliseconds while transformation to bidentate complexes occurred on a time
scale of seconds. This provides further evidence that equilibration times on the order of
hours to days are due to slow diffusion rather than complex formation.
Coprecipitation experiments producing equivalent amounts of ferrihydrite as
45
the above showed much higher adsorption rates (pH 7.5-9.0), with >95% As[V] adsorbed
after 5 min and approximately twice as much As[V] adsorbed after 24 hours compared
to PSA experiments. After the initial uptake, the coprecipitate began to release As[V],
continuing after 750 hours of reaction, though the final amount adsorbed was still greater
than in the PSA case. Fe 2+ derived coprecipitates generally had less initial sorption than
similar systems with Fe 3+ derived coprecipitates, however they tended to desorb more
slowly with aging. EXAFS analysis on PSA and coprecipitation samples revealed no
surface precipitate or solid-solution formed, pointing to diffusion as the rate limiting
step (Waychunas et al., 1993; Fuller et al., 1993).
Numerous researchers have chosen equilibration times for adsorption batch ex-
periments with either no reported explanation or no supporting data. Equilibration times
used for As[V] and As[III] on goethite include 24 hours (Grossl et al., 1997; Hingston
et al., 1971; Dixit & Hering, 2003), 16 hours (Manning et al., 1998), 2 hours (Sun &
Doner, 1996), and overnight (Farquhar et al., 2002). Equilibration times used for As[III]
and As[V] on ferrihydrite and amorphous-iron oxide include 24 hours (Jain et al., 1999;
Raven et al., 1998), 4 hours (citing Pierce & Moore (1982); Dixit & Hering (2003)), 2
hours (Hsia et al., 1992; Hering et al., 1996; Wilkie & Hering, 1996) and 20 minutes
(Goldberg & Johnston, 2001). Equilibration for coprecipitation experiments with ferric
salts include 4 hours (Roberts et al., 2004), 1 hour (citing internal kinetics tests; Meng
et al. (2000)) and a combination of 1 min rapid mixing, 30 min slower mixing, and 15
min quiescent settling (Hering et al., 1996). Dixit & Hering (2003) used 24 hours equi-
libration for As[III] adsorption onto magnetite, citing internal kinetics tests (no details
46
given).
The kinetics of arsenic removal via ZVI depends on the rates of several subpro-
cesses, including the corrosion rate of Fe(0) (which depends on the DO concentration
and pH), the rate of formation of ferric hydroxide precipitates, as well as the rate of
arsenic adsorption onto the precipitates. Bang et al. (2005a) reports a steady state of
soluble iron formation from Fe(0) and ferric hydroxide formation after only 10 minutes
in solution. Bang et al. (2005b) was able to fit a psuedo first-order reaction obtaining
removal half lives for As[III] and As[V] via ZVI at pH 6 of 3.96 and 2.14 hrs respectively.
At pH 7, the half life for As[V] increased to 8.72 hours (half-life of As[III] could not be
measured from the data at pH 7). Su & Puls (2001) cite half-lives of 9-131 hours for
As[V] removal on a variety of commercially available Fe(0) products. Farrell et al. (2001)
and Melitas et al. (2002b) described the arsenate removal kinetics by corroding iron in
batch tests using a model of the form:
d[As]dt
=−k0[As]
k0/k1 + [As](2.8)
where [As] is the arsenate solution concentration, t is time, k0 is the zeroth-order rate
constant, and k1 is the first-order rate constant. Removal rates are first-order in [As]
at the limit of infinite dilution and zeroth-order in [As] at infinite concentration. First-
order kinetics occur in situations with excess adsorption sites (and hence no sorptive
competition), limited by mass-transfer to adsorption sites. Deviations begin as the rate
of Fe 2+ generation becomes less than the rate of removal, inducing competition between
species for a decreasing number of adsorption sites. At a fixed corrosion rate, competition
increases with arsenic concentration. At high arsenic concentrations, removal kinetics
47
may approach zeroth-order, limited by the rate of adsorption site generation (Melitas
et al., 2002b).
It has been reported that arsenate reacts with amorphous iron oxides much
faster than arsenite; Pierce & Moore (1982) found arsenite adsorption onto pre-formed
am-Fe(OH)3 to be 90% complete after 2 hours (final pH after 24 hours 6.1 and 9.8), while
arsenate adsorption was 90% complete after only 1 hour (final pH after 24 hours 8.0 and
9.9). Adsorption of both anions was 99% complete after 4 hours (initial As concentration
of 0.150 mol As/kg am-Fe(OH)3). However, Raven et al. (1998) found arsenite adsorbed
faster to pre-formed ferrihydrite than arsenate at pH 9.2 (maintained throughout reac-
tion), reaching 99.96% of the uptake at 96 hours after 1 hour compared to 99.69% for
arsenate after 1 hour (initial As concentration of 0.267 mol As/ kg Ferrihydrite). Both
Raven et al. (1998) and Pierce & Moore (1982) found noticeably higher adsorption rates
for higher initial arsenic concentrations (13.3 mol As/kg Ferrihydrite and 2.9 mol As/kg
am- Fe(OH)3 respectively) which could help to explain the discrepancy between results.
Fuller et al. (1993) has challenged the results of Pierce & Moore (1982) based on their
failure to maintain a constant pH throughout the reaction time. Bang et al. (2005b)
found As[III] removal to be slower than As[V] removal at pH 7 using ZVI under an oxic
environment, reaching only 55% for As[III] compared to 72% for As[V] after 90 minutes
of mixing.
48
2.4.6 Effects of co-occuring solutes
Many inorganic and organic aqueous species can affect arsenic adsorption. They
can compete with arsenic directly for available binding sites or indirectly by altering the
electrostatic charge at the adsorbent surface. Both processes are influenced by pH,
ion concentration, and the intrinsic binding affinity of the adsorbent (Davis & Kent,
1990). Arsenic affected (As >10 ppb) groundwater in Bangladesh is known to contain
phosphate, silica, sulfate, bicarbonate, calcium, and magnesium among others (see Table
3.1), each of which may affect arsenic sorption onto iron (hydr)oxides in different ways.
Phosphate
Spectroscopic studies have shown than phosphate (PO 3 –4 ) forms identical sur-
face species to arsenate (Waychunas et al., 1993; Hiemstra & Van Riemsdijk, 1999). The
affinity of P[V] for surface sites is also similar (Hingston et al., 1971; Manning & Gold-
berg, 1996a,b; Ryden et al., 1987; Gao & Mucci, 2001; Liu et al., 2001). Jain & Loeppert
(2000) studied the influence of P[V] on adsorption of As[III] and As[V] on ferrihydrite
as a function of pH for P[V]:Astot ratios of 1:1 and 10:1 - within the range reported for
many groundwaters, including those in Bangladesh (BGS, 2001). They found that ad-
sorption of both As[V] and As[III] decreased with increasing P[V] concentration, and for
As[V] this was significant over the entire pH range. At pH 7, As[V] adsorption decreased
from about 96% to 50% between no P[V] and the highest concentration (P[V]:Astot =
10:1). At the same pH, As[III] adsorption decreased from about 96% to 74%. At pH
9, the decrease in As[III] adsorption was only a few percent, even at the highest P[V]
49
concentration, indicating that the neutral H3AsO 03 dominant at this pH is better able
to compete with negatively charges phosphate species. Other researchers have found
similar effects. Hingston et al. (1971) measured an 85% decrease in As[V] adsorption by
goethite when the P[V]/As[V] ratio was increased from 0 to 12:1. Manning & Goldberg
(1996a) reported that a P[V]/As[V] ratio of 1:1 caused a 30% decrease in As[V] adsorp-
tion by both goethite and gibbsite compared to P[V]-free solutions at pH values less than
8. It is important to note that the above studies looked at competitive adsorption on
iron (hydr)oxides formed prior to the addition of arsenic and other ions.
The order in which phosphate and arsenic are added to solution is an important
factor in determining competitive effects. Liu et al. (2001) found an adsorption density
of 78 mmol As[V]/kg goethite when equal amounts of As[V] and P[V] were added to
solution at the same time (at pH 7) compared to only 51 mmol As[V]/kg goethite when
the P[V] was added 24 hours before As[V] and 139 mmol As[V]/kg goethite when As[V]
was added alone. Some researchers have also reported desorption of arsenic on iron
oxides in the presence of phosphate (Alam et al., 2001; Jackson & Miller, 2000).
The structure and particle size of iron (hydr)oxides can also be affected by phos-
phate. Mayer & Jarrell (2000) found that crystallite growth was slower in the presence
of phosphate. Lytle & Snoeyink (2002) found that the particle size of iron oxide sus-
pension, as observed through photon correlation spectroscopy, decreased in the presence
of phosphate. In addition, polyphosphates are commonly added to drinking water to
prevent iron from precipitating out of solution, though the more stable orthophosphate
form is more likely to be present in natural waters.
50
Silicate
The iron (hydr)oxide surface has a strong affinity for silicate (Sigg & Stumm,
1981), however silicate adsorption is weaker than As[V] adsorption (Swedlund & Web-
ster, 1999). Silicic acid (H4SiO4) has been shown to effectively compete with arsenic
for adsorption sites (Davis et al., 2001). The adsorption of As[V] was found to decrease
with increasing pH and silicate concentration (Holm, 2002) and Waltham & Eick (2002)
reported that silicate increased iron mobilization and interfered with arsenic removal.
Meng et al. (2000) looked at Si and As adsorption onto ferrihydrite as ferrihydrite pre-
cipitated from solution. Adsorption of As[III] (300 µg/L) and As[V] (500 µg/L) at pH
6.8 began to decrease in the presence of 1 mg/L Si; at 10 mg/L Si, adsorption decreased
by 70% for As[V] and 80% for As[III]. Ferrihydrite that had been aged for 18-24 hours
prior to adding 3.1 mg/L Si showed little effect on adsorption of As[V] and As[III] at
pH <8 (Swedlund & Webster, 1999). In the presence of 56 mg/L Si, As[III] adsorp-
tion decreased between pH 4 and 10 by as much as 35%, and As[V] asorption decreased
by as much as 60% above pH 6. Swedlund & Webster (1999) attributed the decrease
in arsenic adsorption to surface site competition and polymerization of Si resulting in
negative surface charge.
Silicate has also been known to affect the structure of iron (hydr)oxides. Silicate
can affect the particle size and crystallinity of iron (hydr)oxides (Rushing et al., 2003;
Davis et al., 2001; Meng et al., 2000; Deng, 1997; Anderson & Benjamin, 1985). Mayer
& Jarrell (2000) found that the presence of Si during Fe[II] oxidation at molar ratios
above 0.36 Si/Fe promoted the formation of ferrihydrite, though it inhibited coagulation
51
and crystallite growth. Anderson & Benjamin (1985) similarly found the presence of
silicate delayed ferrihydrite crystallization into goethite from 24 hr to 1-2 weeks. Davis
et al. (2001) found that silicate sorption to iron oxides produced a highly negative surface
charge, leading to the creation of smaller iron particles from larger particles. The effects
of Si on iron (hydr)oxides mineralogy may be more important than the competition
between Si and As for sorption sites. Increased sorption capacities of Fe oxides for
phosphate (similar in structure to arsenate) in the presence of silica has been attributed
to the effect of Si on Fe oxide mineralogy as opposed to competition between P and Si
for sorption sites (Mayer & Jarrell, 2000). However, the increased sorption capacity for
phosphate could also increase the competitive edge of phosphate compared to arsenic in
sorption site competition, enhancing the detrimental effect of phosphate.
There has been some evidence that silica can catalyze the oxidation of Fe[II] to
Fe[III] in aerated water in the pH range of 6.6 to 7.1 (Schenk & Weber, 1968). However,
Rushing et al. (2003) found that silica significantly slowed Fe[II] oxidation to Fe[III] in
aerated water at pH 6.5.
Carbonate
Carbonate species exist in natural waters as H2CO 03 , HCO –
3 , and CO 2 –3 . There
is some evidence of carbonate species adsorption onto iron (hydr)oxides (Villalobos &
Leckie, 2000; van Green et al., 1994; Zachara et al., 1987). It is usually described as
an outer-sphere complex, but spectroscopic evidence suggests the formation of a mon-
odentate inner-sphere complex (Villalobos & Leckie, 2000; van Green et al., 1994). Meng
52
et al. (2000) found essentially no effect of carbonate on adsorption of As[III] and As[V] on
ferrihydrite in equilibrium with air and with a total carbonate concentration of 0.01mM.
Little effect was found on adsorption of As[V] during coprecipitation in stream water
with 3mM carbonate at pH 8 (Fuller et al., 1993). There is some evidence that carbon-
ate may affect adsorption of As[III] but not As[V]. Wilkie & Hering (1996) found that
1mM carbonate caused about a 10% decrease in As[III] adsorption on ferrihydrite at pH
9 and only a few percent decrease in As[V]; no effect was seen at pH 6. Some evidence
also exists that adsorption of carbonate by ferrihydrite in a shallow aquifer (5mM car-
bonate, 7.1 >pH >7.7) may have decreased adsorption of As[III], and to a lesser extent
As[V] (Appelo et al., 2002). High arsenic concentrations in groundwater have also been
correlated with high bicarbonate concentrations (Nickson et al., 2000; Kim et al., 2000).
Calcium and Magnesium
Some solutes can reportedly enhance adsorption of arsenic. Wilkie & Hering
(1996) found an increase in As[V] adsorption by ferrihydrite when calcium was also
adsorbed at pH 9, apparently by increasing the positive charge on the surface. Meng et al.
(2000) also noted that additions of calcium and magnesium to ferrihydrite suspensions
negated part of the competitive effect of silica on arsenic adsorption. The presence of
calcium has also enhanced adsorption of As[V] on aluminum oxides at pH >8 (Ghosh &
Teoh, 1985).
53
Sulfate
Macroscopic chemical evidence suggests that sulfate [S(VI)] adsorbs via an
outer-sphere complex and should therefore have much less effect on arsenic adsorption
than inner-sphere complex formers such as phosphate (Stollenwerk, 2003). Jain & Loep-
pert (2000) studied the influence of sulfate at S/As ratios of 10:1 and 50:1 on As[III] and
As[V] adsorption to ferrihydrite from pH 3-11. They found essentially no effect on As[V]
adsorption over the entire range, though an increase of a few percent was observed for
the highest S/As ratio, attributed to an ionic strength effect. There was a slight decrease
in As[III] adsorption at pH <7, but no effect at higher pH. Similar results were observed
by Wilkie & Hering (1996). As[III] adsorption by ferrihydrite decreased 20% at pH 7
and 70% at pH 4 in the presence of 4000:1 S/As, while As[V] adsorption decreased by
only a few percent at pH 5.
Combined effect of ions
Meng et al. (2002) looked at the combined effect of phosphate, silicate, and bi-
carbonate on the removal of arsenic from Bangladesh groundwater and simulated ground-
water by iron (hydr)oxides. They found that phosphate (0-0.08 mM), silicate (0-0.8 mM),
and bicarbonate (0-14 mM) in separate solutions had none to moderate effects on As[V]
removal in solutions with 6.7 mg/L Fe and 300 ppb As[V]. However, the presence of
bicarbonate and silicate was found to magnify the adverse effect of phosphate such that
the residual As[V] concentration increased 10 fold. Roberts et al. (2004) investigated the
amount of Fe[II] and Fe[III] required in coprecipitation experiments to reduce 500 ppb
54
of As[III] and As[V] to below 50 ppb with and without 30 mg/L Si and 3 mg/L P. Over
2.5 and 7 times the Fe[III] was needed for As[III] and As[V] respectively in the presence
of Si and P. Using Fe[II] additions, about 1.5 and 5 times as much was needed for As[III]
and As[V] respectively. Mayer & Jarrell (2000) found that the presence of silica during
Fe[II] oxidation greatly increased the sorption density and retention capacity of the re-
sulting ferrihydrite for phosphate. This could potentially increase the detrimental effect
of phosphate on arsenic adsorption if the affinity of iron (hydr)oxides for phosphate is
increased more than the affinity of iron (hydr)oxides for arsenate or arsenite.
The presence of co-occuring solutes can also affect the relative efficiency of
arsenic removal via coprecipitation with Fe 2+ and Fe 3+ salts. Roberts et al. (2004)
observed that for synthetic groundwater free of P and Si, As[V] removal was similar
using either Fe 2+ or Fe 3+, however in the presence of 30 mg/L Si and 3 mg/L P, both
As[V] and As[III] were more efficiently removed by the addition of Fe 2+, with the effect
being more pronounced for As[III].
2.5 Electrocoagulation
2.5.1 Electrochemical cell theory
To help follow the rest of the thesis with minimal reference to external texts,
below is a summary of published material heavily based on Bard & Faulkner (2001), but
also embellished with additional references, inferences, and recorded observations.
A simple electrochemical cell consists of two electrical conductors (electrodes)
55
separated by an ionic conductor (electrolyte). Charge is transported in the electrodes
by the movement of electrons (and holes). Charge is transported in the electrolyte by
the movement of ions. The boundary between two chemical phases, such as the solid
electrode and aqueous electrolyte, is called the interface, and numerous processes and
factors can affect the transport of charge across the interface. The difference in electric
potential between the electrodes is called the cell potential, measured in volts (V), and
is a measure of the energy available to move charge between electrodes. It is the sum of
the electric potential across all of the various phases in the cell. The transition in electric
potential crossing from one conducting phase to another (e.g. electrode to electrolyte)
occurs almost entirely at the interface (Bard & Faulkner, 2001), implying that high
electric fields can exist in this region. The magnitude of the potential difference at the
interface affects the energies of the charge carriers on either side of the interface, and
hence it controls the direction and rate of charge transfer. The potential difference at
each electrode/electrolyte interface controls the half-reaction occurring there, and hence
the real chemical changes at the electrode (in turn controlling the chemical composition
near the electrode). Commonly, one is interested in reactions occurring at only one of
the electrodes, and that electrode is called the working or indicator electrode.
The cell potential is the sum of many potential differences across the cell, includ-
ing each interface between phases and the ohmic drop across the electrolyte. To measure
or control the potential at the working electrode alone, the non-working electrode is of-
ten a reference electrode, made up of components with essentially constant composition
capable of maintaining a constant interface potential over a wide range of currents. The
56
ohmic drop can be minimized by reducing the bulk solution resistance or using a three
electrode cell (discussed in more detail below). The internationally accepted primary
reference is the standard hydrogen electrode (SHE) in which H2 gas is bubbled near a
platinum wire and all components have a unit activity. Standard potentials of a half
reaction are measured in a simple electrochemical cell in which one electrode is a SHE.
Other common reference electrodes include the saturated calomel electrode (SCE) with
a potential of 0.242 V vs. SHE (Bard & Faulkner, 2001) and the silver-silver chloride
electrode with a potential of 0.197 V vs. SHE (Bard & Faulkner, 2001). Control over
the potential of the working electrode is thus achieved only with respect to the constant
potential of the reference.
Driving the potential of the working electrode to more negative values (e.g.
by connecting a battery or power supply to the cell with its negative side attached to
the working electrode) will raise the energy of the electrons in the electrode. They can
reach an energy high enough to transfer into vacant states on a species in the electrolyte,
creating a flow of electrons from electrode to solution known as a reduction current.
Similarly, driving the potential of the working electrode to more positive values can
lower the energy of the electrons in the electrode enough for electrons on solutes in the
electrolyte to transfer to more favorable energy states on the electrode, creating a flow
from solution to electrode known as an oxidation current. The critical potentials at
which these processes occur are related to the standard potentials, E0, for the specific
chemical substances in the system. The standard potential is measured in a cell where all
the constituents are in their standard states (i.e. 1 M at 25C and 1 bar total pressure).
57
Frequently, the concentrations of the reduced or oxidized species are different from 1M.
Consider the reaction:
O + nee− −−−− R (2.9)
O = Oxidized species
R = Reduced species
In this case the critical potential is given by the Nernst equation:
E = E0′ +RT
neFln
[CO][CR]
(2.10)
R = Universal gas constant [kJ/mol-K]
T = Temperature [K]
ne = number of electrons in redox reaction
F = Faraday’s constant [96,485.4 C/mol e−]
CO = Concentration of oxidized species [M]
CR = Concentration of reduced species [M].
Here, E0′ is the formal potential, related to E0 by:
E0′ = E0 +RT
neFlnγOγR
(2.11)
γO = activity coefficient of oxidized species
γR = activity coefficient of reduced species.
The take-away from Equations 2.10 and 2.11 is that the critical potential depends on
environmental conditions, such as temperature, as well as the relative concentrations of
the reactants near the electrode surface and the activity of those reactants.
For faradic processes, the number of electrons that cross an interface is related
stoichiometrically to the extent of chemical reaction (i.e. the amount of reactant con-
sumed and product generated). Current is the total charge passed per unit time, and
58
thus the current is a measure of the rate of chemical reactions occurring in the cell. Un-
der some conditions, a given electrode/solution interface will show a range of potentials
where no charge-transfer reactions occur because they are thermodynamically or kineti-
cally unfavorable. However, adsorption or desorption reactions can occur that change the
structure of the electrode-solution interface without changing the potential or solution
composition. These are called non-faradic processes and can sometimes lead to tran-
sient currents as the electrode area, potential, or solution composition changes. These
currents are not associated with charge-transfer. When one plots the current as a func-
tion of potential, one obtains a current-potential (i vs. E) curve including information
about both faradic and non-faradic processes. Such curves can be very informative about
the reactions occurring at the electrode (this is the basis of the transient voltammetry
technique used in Chapter 3).
If well-defined redox couples exist at each electrode, then equilibrium can be
established and the cell will have a well defined equilibrium potential, or open circuit
potential, Eeq. This is the potential one would measure across the electrodes if no net
current was flowing. In many cases, there is no well-defined equilibrium state for the
cell, and the open circuit potential can only be placed within a potential range. The
departure of the electrode potential from the equilibrium value upon the passage of a
faradic current is termed polarization. The extent of polarization is measured by the
overpotential, η:
η = E − Eeq (2.12)
The overpotential η can be considered a sum of terms associated with different reaction
59
steps:
η = ηMt + ηCt + ηrxn (2.13)
ηMt = mass-transfer overpotential [Volts]
ηCt = charge-transfer overpotential [Volts]
ηrxn = the overpotential associated with a preceding reaction [Volts]
ηMt is the overpotential necessary to drive mass-transfer, the physical movement of ions
from the bulk solution to the electrode surface where reactions take place (also called
the concentration overpotential or diffusion overpotential). It is caused by the change
in solute concentrations occurring in the proximity of the electrode surface due to the
electrode reaction. The mass-transfer overpotential can be reduced by increasing the
transport of ions from anode surface to the bulk solution, i.e. by increased mixing
between electrodes. ηMt becomes negligible when the reaction rate constant is much
smaller than the mass-transfer coefficient (Mollah et al., 2004). The charge-transfer
overpotential, ηCt, is the energy associated with electron transfer at the electrode surface
and ηrxn is energy associated with the chemical reactions that precede or follow electron
transfer. The sum of ηCt and ηrxn is sometimes called the kinetic overpotential, ηK .
Both ηK and ηMt increase as the current density, j = i/A, increases. A precise
relationship can be derived in the simple case of a one-step, one-electron process, leading
to the Butler-Volmer formulation. For the case of interface equilibrium and a solution
in which the bulk oxidized species concentration is equal to the bulk reduced species
concentration, the Butler-Volmer formulation is (see Bard & Faulkner (2001), section
3.3.3 for a detailed derivation):
j = Fk0[CO(0, t)e−αfη − CR(0, t)e(1−α)fη] (2.14)
60
j = Current Density
F = Faraday’s constant
k0 = standard rate constant for the reaction
CO(0, t) = concentration of oxidized species at the electrode surface as a function oftime
CR(0, t) = concentration of reduced species at the electrode surface as a function oftime
α = transfer coefficient, ranging from zero to unity
f = F/RT
η = overpotential at the electrode
The standard rate constant, k0 (s−1), is the rate constant for the reaction when η = 0.
The transfer coefficient α is a property of the system. Equation 2.14 demonstrates the
tendency for the current density to be related exponentially to the overpotential η. This
behavior is seen in many electrochemical systems, both simple and complicated, as first
noted by Tafel in 1905.
The applied cell potential necessary to get the desired current must take into
account the potential across the working electrode, E (which includes the overpotential),
as well as the voltage drop across the solution due to the bulk solution resistance, Rs.
Using the convention of positive current, i, for oxidation (or anodic) current 1, the applied
cell potential, Eappl is:
Eappl = Eeq + η + iRs (2.15)
The solution resistance, Rs (Ω), is determined by:
Rs =d
A · κ(2.16)
1Note that this is opposite the convention used in Bard & Faulkner (2001)
61
d = distance between electrodes [m]
A = active surface area of anode [m2]
κ = specific conductivity of bulk solution [103 mS/m]
The bulk resistance Rs, and hence the iRs-drop, can be reduced by decreasing the dis-
tance between electrodes, increasing the submerged surface area of anode, or increasing
the specific conductivity of the bulk solution.
For precise measurements of the working electrode interface potential in low
conductivity solution, a three-electrode cell is often used. In this configuration, current
is passed between the working electrode and a counter electrode. The counter electrode
can be any convenient metal, because its electrochemical properties do not affect the
behavior of the electrode of interest, though care must be taken that it will not produce
solutes that can reach the working electrode and cause interfering reactions. A reference
electrode is added with its tip very close to the working electrode to minimize the iRs-
drop. The working electrode potential is measured using a high impedance voltmeter
between the working electrode and the reference electrode, allowing almost all of the
current to pass through the cell from working to counter electrode, and negligibly small
current to pass between the working and reference electrode. Even in this case, if the
reference electrode is not touching the working electrode (it must be at least 2d away
from the working electrode to avoid shielding errors if its tip diameter is d), some fraction
of iRs, known as the uncompensated resistance, will still be measured on the voltmeter.
62
2.5.2 Electrocoagulation (EC) using iron
Electrocoagulation (EC) is a complicated process in which consumable elec-
trodes supply ions that can be used in situ as coagulants or sorbates. When the elec-
trodes of an electrochemical cell are connected to an electrical DC power source, the
pull of electrons from the anode will oxidize the anode metal, causing metal cations to
dissolve into solution. In the case of an iron anode, the oxidized metal ions quickly
hydrolyze to form insoluble polymeric iron (hydr)oxides, providing an active surface for
sorption of contaminants, as well as excellent coagulating agents. Coagulation occurs
when the metal cations combine with negative particles carried towards the anode by
electrophoretic motion (Mollah et al., 2004). The negative ions neutralize ionic iron
species in the solution, reducing electrostatic interparticle repulsion until van der Waals
attraction predominates, aiding coagulation and aggregation into flocs (Mollah et al.,
2004). The flocs create a sludge blanket able to entrap and bridge colloidial particles
(Mollah et al., 2004), removing contaminants while enhancing further aggregation. Con-
taminants can precipitate directly or physically and chemically attach to the aggregating
iron (hydr)oxides. This allows for the removal of colloidal contaminants via filtration or
sedimentation of the aggregated iron (hydr)oxides they are attached to. In addition to
the processes described above, some of the following may also occur in the electrocoag-
ulation cell, which may hinder or aid contaminant removal (Mollah et al., 2004):
• Cathodic reduction of impurities.
• Anodic oxidation of impurities.
• Discharge and coagulation of colloidal contaminant particles.
63
• Electrophoretic migration of ions in solution, increasing the probability of collisions(Persin & Rumeau, 1989).
• Electroflotation of coagulated particles by O2 and H2 bubbles produced at theelectrodes (bubbles can float contaminant flocs to the surface).
• Reduction of metal ions at the cathode.
• Other electrochemical and chemical processes.
The production of iron (hydr)oxides from the electrochemical dissolution of an
iron anode may proceed through a number of different chemical reactions. Oxidation at
the anode can produce Fe 2+ or Fe 3+ according to either:
Fe(s)−−−− Fe 2+
(aq) + 2 e− (2.17)
or
Fe(s)−−−− Fe 3+
(aq) + 3 e−. (2.18)
For Faradic processes, the theoretical amount of Fe 3+ or Fe 2+ ions, wth (mg), dissolved
into solution at current i in current processing time tCP is given by Faraday’s law:
wth =i · tCP ·M
nF(2.19)
i = operating current (A)
tCP = current processing time (s)
M = molecular weight (for iron, M = 55.845 g/mol)
n = number of electrons involved in the oxidation/reduction reaction (n = 2 or 3 forFe 2+ or Fe 3+ respectively)
F = Faraday’s constant. 96,500 C/mol e−
64
Equation 2.19 assumes that every coulomb passing through the anode is used
to oxidize iron. In reality, some of the charge often goes into oxidizing other available
ions (side reactions), such as the oxygen evolution via:
2 H2O(l)−−−− O2(g) + 4 H + + 4 e− (2.20)
The current efficiency, φ, is often invoked to describe the fraction of coulombs supplied
that are actually used to oxidize iron:
φ =wexpwth
(2.21)
where wexp (mg) is the measured weight of iron dissolved. wexp is determined experi-
mentally by measuring the final iron concentration in solution or the difference between
electrode weight before and after applied voltage. Both methods fail to take into account
the buildup of an active oxide layer on the electrode surface which does not break into
solution. An active layer on the electrode surface will lead Equation 2.21 to underesti-
mate the actual iron dissolved by an amount proportional to the thickness of the layer,
thereby slightly underestimating the current efficiency. Assuming that φ is accurate, the
actual iron dissolved w becomes:
w = wth · φ (2.22)
Once in solution, the iron ions immediately hydrolyze (within milliseconds) to
form (depending on pH) a combination of FeOH+, Fe(OH) 02 , and Fe(OH) –
3 in the case
of Fe 2+, or FeOH 2+, Fe(OH) +2 , Fe(OH) 0
3 , and Fe(OH) –4 in the case of Fe 3+. Fe[II]
species may either precipitate out as Fe[II](hydr)oxides and FeCO3 (s) or oxidize and
precipitate to form Fe[III](hydr)oxides. Fe[III] species may precipitate directly into
65
Fe[III](hydr)oxides. The exact combination of end products is a sensitive function of
the potential, pH, and concentration of various species in solution. The resulting phase
may be a meta-stable amorphous or freshly-precipitated crystalline mineral (Arienzo
et al., 2002). The iron (hydr)oxides produced are sparingly soluble at neutral pH (e.g.
HFO, ksp = 10−38 (Martin & Kempton, 2000).
The complete generation of iron (hydr)oxides by an iron anode in electroco-
agulation is often summarized by the following two mechanisms (Mollah et al., 2004;
Daneshvar et al., 2007; Yildiz et al., 2007; Endyuskin et al., 1983):
• Mechanism 1
4 Fe(s)−−−− 4 Fe 2+
(aq) + 8 e− (2.23)
4 Fe 2+(aq) + 10 H2O(l) + O2(g)
−−−− 4 Fe(OH)3(s) + 8 H +(aq) (2.24)
Cathode:
8 H +(aq) + 8 e− −−−− 4 H2(g) (2.25)
Overall:
4 Fe(s) + 10 H2O(l) + O2(g)−−−− 4 Fe(OH)2(s) + 4 H2(g) (2.26)
• Mechanism 2:
Anode:
Fe(s)−−−− Fe 2+
(aq) + 2 e− (2.27)
Fe 2+(aq) + 2 OH−(aq)
−−−− Fe(OH)2(s) (2.28)
66
Cathode:
2 H2O(l) + 2 e− −−−− H2(g) + 2 OH−(aq) (2.29)
Overall:
Fe(s) + 2 H2O(l)−−−− Fe(OH)2(s) + H2(g) (2.30)
In both mechanisms, hydrogen evolution occurs at the cathode and Fe(OH)3 is the as-
sumed iron (hydr)oxides end product. A schematic representation of the EC arsenic
removal process is shown in Figure 2.1. As stated above, the type of iron (hydr)oxides
actually produced at the electrode is a sensitive function of potential, pH, and concen-
tration of various species in solution. A Pourbaix diagram is often used to show the
dominant species in equilibrium as a function of potential and pH for a given aqueous
concentration. Figure 2.2 shows a simplified Pourbaix diagram for iron (assuming only
iron is present in solution) at a given aqueous iron concentration. To demonstrate the
complexity of a real system, imagine the entire Pourbaix diagram shifting slightly as the
concentration of aqueous iron increases through continuous dissolution of the electrode
(Equation 2.19). Further imagine that the electrode potential changes (Equation 2.10) as
concentrations of reduced and oxidized species shift. The pH near the electrode may also
change as electrochemical and chemical reactions occur and OH – ions produced at the
cathode migrate to the anode in the applied electric field. Recall also that the Pourbaix
diagram represents thermodynamic equilibrium, which may or may not be reached at a
given potential depending on the rates of charge-transfer, mass-transfer, and chemical
reactions in the system. Solids can precipitate and exist for hours to hundreds of years
without reaching thermodynamic equilibrium (meta-stable states). Finally, recall that
67
the Pourbaix diagram is a function of the entire solution composition, including other
co-occuring solutes that may or may not be electrochemically active at a given poten-
tial. For all of these reasons, the dominant type of iron (hydr)oxides species produced is
extremely difficult to predict.
All major iron (hydr)oxides have been identified in the corrosion products of
iron and steel (Cornell & Schwertmann, 2000). In some cases, the physical placement
of corrosion products is more or less random, while in others, the different oxides are
arranged in layers. Layer-type rust results from potential or chemical gradients across
the oxide film. Such gradients often change with film thickness, leading to rust compo-
sition changes with distance from the metal (Cornell & Schwertmann, 2000). Arsenic
adsorption onto ZVI is thought to primarily occur on the oxide film forming around iron
filings (Melitas et al., 2002b). In EC, it is not obvious whether arsenic removal is due
to adsorption to an iron (hydr)oxides film formed on the anode (which may or may not
break off into solution), or adsorption to iron (hydr)oxides formed in solution. Pinisakul
et al. (2002) recovered 65-85% of spiked arsenic from the sludge present in bulk solution
after electrocoagulation experiments (initial pH 3-10.5) and only 0.03-1.1% was found
adsorbed to the electrode plates and reactor walls, suggesting that adsorbed arsenic does
not remain on the electrode. The iron (hydr)oxides produced will also age on timescales
of minutes to days, tending to become more crystalline (see Section 2.3). Thus adsorp-
tion may occur on slightly more aged iron (hydr)oxides towards the end of electrolysis
that at the beginning.
68
Figure 2.1: Schematic representation of the removal of arsenate ions from solution viaEC. For this schematic, Fe 3+ ions are dissolving from the electrode, though Fe 2+ ionsmay dissolve as well. In addition, Fe(OH)3 is used to represent the precipitating iron(hydr)oxides, though the actual composition of iron (hydr)oxides formed may be quitecomplex (see text).
69
Figure 2.2: Pourbaix diagram for iron. Assumes iron is the only species present. Eh isthe equilibrium potential relative to a standard hydrogen electrode (adapted from Hem(1961))
70
Though the exact composition of iron (hydr)oxides end products is difficult to
predict, the overpotential η at the anode/solution interface is clearly one of the controlling
variables. The overpotential can be adjusted by changing the applied cell potential Eappl
(Equation 2.15). However, in EC it is often desirable to run in galvonostatic mode, in
which the current is kept constant through continuous small adjustments of Eappl. In this
case, the overpotential is controlled by the current density j via Equation 2.14. Some
drifting of η may occur during the process to maintain a constant current as CO(0, t)
and CR(0, t) evolve with time, potentially leading to a shift in the thermodynamically
favorable composition of ironox during the process. However, since j is exponentially
dependent on η, but linearly dependent on concentration, the drift is likely to be relatively
small.
Two important parameters used to control EC are thus the current density j,
which determines the composition of iron (hydr)oxides end products, and the iron dosage
w which determines the quantity of iron (hydr)oxides end products. As seen in Equations
2.19 and 2.22, w is determined by the operating current i, the current processing time tp,
and the current efficiency φ. Since φ can only be determined experimentally, the charge
density q is often cited instead of w, where q (C/L) is given by:
q =itCPV
(2.31)
i = operating current (A)
tCP = current processing time (s)
V = active solution volume (L)
71
If the active solution volume changes during the current processing time, for example as
aliquots are removed for sampling, the charge density must be calculated using:
q =∑i
i∆tCP iVi
(2.32)
i = operating current (A)
∆tcpi = duration of current processing time during which the active volume is Vi (s)
Vi = active solution volume for duration ∆tpi (L)
where the total current processing time is:
tCP tot =∑i
∆tCP i. (2.33)
The processing time tCP is often considered important only to the extent that
it affects q. However, it is possible for tCP to affect the EC process independently of q if
coagulation and aggregation are significantly enhanced by the EC process. In this case,
more arsenic could be removed per mg of iron (i.e. per unit of q) when tCP is longer
simply because larger particles are more easily separated from the water than smaller
ones. In addition, a longer processing time gives the arsenic more time to complex with
the continuously generated iron (hydr)oxides, which can be significant if the water is
filtered immediately after EC treatment (i.e. with no additional complexation time after
the current is turned off). Due to these potential effects, tCP should be considered an
independently significant operating parameter controlling the EC process, along with j
and q.
The electrode surface area to treatment volume ratio, A/V , could also be an
important controlling EC variable. The electrode area influences current density, posi-
72
tion of cation dosing, as well as bubble production and bubble path length (important
parameters for electroflotation). However, if filtration is used in place of electroflotation
(making the bubble efficiency less important), and the solution is thoroughly and rapidly
mixed (leading to even cation dosing regardless of the electrode size and position), then
the electrode area should only effect the arsenic removal capacity (mg-As-removed/C) to
the extent that it affects current density. Thus, with respect to arsenic removal capacity,
it should not be considered a controlling EC variable separate from the current density.
Note however that the A/V ratio could separately affect the time required to achieve a
given dosage q. Rearranging the charge density equation (2.31) in terms of j and solving
for tCP , one gets:
tCP =q
j(A/V )(2.34)
Thus, although A/V may not significantly affect the quantity of charge q required to
remove n µg/L of arsenic (the arsenic removal capacity), it may affect the amount of
time tCP required to achieve a given dose q. This effect was noted by Mameri et al.
(1998) while studying the effect of A/V on fluoride removal from Saharan groundwater
using aluminum electrodes (a similar process in which removal is based on adsorption
of fluoride to electrochemically generated aluminum hydroxides). They reported that
minimum processing time required to reduce fluoride concentrations a given amount at
a fixed current density decreased as A/V increased, exactly the relation expected by
Equation 2.34. They also reported that beyond a value of A/V = 28.7 m2/m3, no
significant effect of A/V on the processing time was observed, suggesting that additional
effects may come into play (in the case of Mameri et al. (1998), these additional effects
73
could be due to the use of electroflotation).
Equation 2.34 has additional significance as well. It points out that the EC
operating parameters q, j, and tCP , each of which may have distinct control over the
arsenic removal capacity, are not completely independent of each other. They can only
be adjusted independently if A/V is used as a compensatory variable. A/V in turn is
limited by the physical constraints of the electrochemical cell and the extent to which
the assumption that A/V does not affect the arsenic removal capacity is valid.
2.5.3 As[III] oxidation in EC
One of the major advantages of using EC to remediate arsenic is the potential
for a net oxidation of As[III] species in the water to As[V], which binds more readily to
the ironox produced by corrosion at neutral pH (see Section 2.4.4). Water treatment
techniques using chemical addition frequently have to add an oxidizing step before re-
moving As[III] (Arienzo et al., 2002), which can degrade water quality and add to cost.
The equilibrium potential for the As[III]/As[V] redox couple depends on the relative
concentrations of As[III] and As[V]. It can also depend on whether the reaction occurs
to bound or aqueous arsenic species (Melitas et al., 2002a). The reaction will take place
very close to the electrodes, where the pH can be quite different from the bulk solution
due to other reactions taking place there and the migration of OH – ions produced at
the cathode. Oxidation of arsenic at the cathode can proceed according to:
H3AsO3 + H2O −−−− H3AsO4 + 2 H + + 2 e− (2.35)
74
Reduction of As[V] back to As[III] can occur at the cathode as well. If the cathode
is not separated from the anode, net oxidation can only occur if oxidized As[III] ions
adsorb to ironox as As[V] before migrating to the cathode. Since As[V] tends to exist as
negatively charged oxyanions at neutral pH (see Section 2.2), it will tend to drift away
from the cathode, aiding net oxidation of As[III]. Predominant As[III] species are neutral
at neutral pH will not be pulled by the applied electric field.
As[III] oxidation could also occur in parallel with Fe[II] oxidation due to dis-
solved O2 (see Section 2.4.3 for more details). Hug & Leupin (2003) have found this
reaction to contribute to As[III] removal via zero-valent iron (ZVI).
Net oxidation of As[III] during EC was inferred by Kumar et al. (2004) who
measured an initial increase in As[V] concentration during EC on Indian tap water spiked
only with As[III]. They attributed the higher removal of As[III] seen in EC relative to
the chemical addition of Fe[III] salts to As[III] oxidation. Gomes et al. (2007) verified
some oxidation of As[III] to As[V] during EC using X-ray Photon Spectroscopy (XPS).
Pinisakul et al. (2002) inferred oxidation of As[III] to As[V] due to the presence of
Fe4As2O11 in precipitated sludge when only As[III] was present in the original solution.
One research group measured AsH3 gas produced at the cathode during elec-
trocoagulation (Pinisakul et al., 2002). The presumed cathodic reaction was:
HAsO2 + 6 H + + 6 e− −−−− AsH3 (g) + 2 H2O. (2.36)
Between 11-16% of the arsenic removed was attributed to formation of AsH3 gas. The
redox equilibrium potential for Equation 2.36 is E0 = −0.35V relative to a standard
hydrogen electrode, only slightly less negative than the equilibrium potential of hydro-
75
gen evolution (Equation 2.25) with E0 = −0.88V (Pinisakul et al., 2002). No other
investigations reviewed here have attributed arsenic removal via EC to the formation of
arsine gas.
2.5.4 Arsenic removal mechanism using EC
The mechanism of arsenic removal in EC is to produce complexes with arsenic
that are large enough to be separated from water through either filtration or sedimen-
tation and decantation. The mechanism of complex formation is not completely known.
One possible process is adsorption of arsenic oxyanions to electrochemically generated
iron (hydr)oxides. Colloidial arsenic-iron (hydr)oxides complexes may form and then
aggregate, or arsenic may adsorb to surface of aggregated iron (hydr)oxides. A second
possible process is the precipitation of an insoluble arsenic-iron phase (such as ferric
arsenate). Surface precipitation of an arsenic-iron phase could also occur on the surface
of aggregated iron (hydr)oxides. Other processes that may occur include electrochemi-
cally enhanced coagulation of colloidal arsenic oxyanions or even occlusion, which is the
entrapment of contaminants in the interior of growing particles. As[III] removal may be
enhanced by oxidation to As[V] followed by complexation.
2.5.5 Arsenic removal studies using EC
EC has been studied in bench scale reactors to remove arsenic from wastewater
(Hansen et al., 2007, 2006; Balasubramanian & Madhavan, 2001; Gomes et al., 2007) and
potable water (Pinisakul et al., 2002; Kumar et al., 2004; Parga et al., 2005b; Arienzo
76
et al., 2002). Wastewater studies have demonstrated the ability of EC to reduce influent
arsenic concentrations of 100 ppm to below 2 ppm at low initial pH (pH 1-2) using mild
steel electrodes with (Hansen et al., 2006) and without (Balasubramanian & Madhavan,
2001) additional air injection to facilitate oxidation of Fe 2+ to Fe 3+. Hansen et al. (2006)
operated at a current density of j = 12 mA/cm2 and (Balasubramanian & Madhavan,
2001) at j = 5 and 12.5 mA/cm2. Gomes et al. (2007) used iron plate electrodes to
reduce 13.4 ppm As[III] to less than 50 ppb in spiked salt water (j = 3 and 30 mA/cm2).
Ahmed Basha et al. (2008) lowered arsenic concentrations of 2000 ppm in acidic (pH < 1)
copper smelting effluent to 100 ppm using EC to reduce arsenic to arsine (AsH3) as well as
reduce co-occuring sulphate to a species that will readily form insoluble arsenic sulphides
(j = 20 and 40 mA/cm2). EC has also been used to reduce arsenic concentrations to
levels acceptable in potable water. Arienzo et al. (2002) used EC with iron and the
addition of 10-500 ppm H2O2 to reduce 2.5 ppm As[III] in spiked tap water to less than
10 ppb for pH 3.5-9.5 (j = 0.4 mA/cm2). Parga et al. (2005a) used EC with carbon
steel and air injection to reduce 2ppm samples to less than 2 ppb (initial pH 2.86) and
successfully reduced arsenic in well water from La Comarca Lagunera, Mexico from initial
values of 25-300 ppb to less than 2 ppb under continuous flow at a rate of 2500 m3/day
(j = 3.7-4.6 mA/cm2). Kumar et al. (2004) reduced 2 ppm of As[V] and As[III] using
EC with iron in spiked Indian tap water to less than 10 ppb at pH 6-8 after about 100
C/L was passed (j = 0.65 and 1.53 mA/cm2). Arsenic removal is generally attributed
to oxidation of As[III] followed by adsorption of As[V] to iron (hydr)oxides in solution.
Different electrode materials were compared by arsenic removal capacity in sev-
77
eral investigations. Gomes et al. (2007) looked at combination Al-Fe electrodes (polarity
reversed every 15 minutes) compared to Fe-Fe and Al-Al electrodes. They found similar
removal capacity for Al-Fe and Fe-Fe electrodes, with significantly less removal capacity
using Al-Al electrodes alone. Kumar et al. (2004) looked at monopolar aluminum, iron,
and titanium anodes and found removal efficiency increased in the sequence aluminum
< titanium < iron. Iron has consistently been found to be the most efficient material
for arsenic removal.
Notably, few of the above studies stress or monitor any additional complexation
time following the electrolysis step in EC. Arienzo et al. (2002) mentions that the post-
electrolysis solution is allowed to settle before filtration, but no time is given. Hansen
et al. (2006) discuss continuous flow reactors that include a basin for precipitates to
settle and cite a distinct current processing time (9.4 minutes) and sedimentation time
(80 minutes), but the settling basin is not separate from the electrolysis chamber. Many
EC studies do not mention the filtration or separation step at all (Balasubramanian &
Madhavan, 2001; Kumar et al., 2004; Parga et al., 2005b; Mameri et al., 1998), though
filtration is implied by the theory stated. In these cases, filtration is assumed to be
immediately following electrolysis. Thus there is no study attempting to separate of the
effect of iron (hydr)oxides dose (controlled by the electrolysis or current processing time,
tCP ) from the effect of arsenic-iron (hydr)oxides complexation time on arsenic removal
efficiency in EC.
78
2.5.6 Kinetics of EC
No study of arsenic removal via EC reviewed here fit data to a kinetic model.
Many studies reported that initial arsenic removal was rapid, followed by a slower phase
(Pinisakul et al., 2002; Kumar et al., 2004; Balasubramanian & Madhavan, 2001). Pin-
isakul et al. (2002) treated initial arsenic concentrations of 0.5-5 ppm and pH 3-10.5.
They report that of the arsenic removed at the end of 20 minutes, 99% had been re-
moved in the first 2 minutes. Kumar et al. (2004) reports 50-60% of the removal seen
after 60 minutes occurred within the first 5 minutes (initial As 2 ppm, pH 6-8). Bala-
subramanian & Madhavan (2001) reported a sharp drop in removal rate that eventually
approaches a constant value (initial As 100 ppm). All attribute the initial rapid removal
to the abundance of arsenic oxyanions available initially compared to later times.
Mameri et al. (1998) obtained good fits for fluoride removal using EC with
aluminum electrodes to a first order model of the form:
dC
dt= −KiC (2.37)
where t, C, and Ki are the time, concentration, and kinetic constant respectively. The
kinetic constant depends on the current density and temperature. This model fit well
over a current density range of 2.89 - 28.9 mA/cm2 at 20C, confirming that defluori-
dation follows an exponential law with time. Mameri et al. (1998) attributed the first
order behavior to the dual phenomena leading to fluoroaluminum complexes in EC: first
the formation of aluminum hydroxide from Al 3+ ions on the electrodes, and second the
formation of fluoroaluminum complexes by fluoride adsorption on aluminum hydroxide.
79
One expects that analogous processes must occur to produce arsenic-iron (hydr)oxides
complexes. However, additional processes may occur as well, including the electrochem-
ical oxidation of As[III] to As[V] and the oxidation of Fe 2+ to Fe 3+.
2.5.7 Characterization of EC products
Several investigators have studied the electrochemically generated by-products
of EC (i.e. the iron (hydr)oxides produced). Pinisakul et al. (2002) used X-ray fluo-
rescence and X-ray diffraction (XRD) to study the dried precipitated sludge produced
from EC treatment at constant voltage (electric gradient of 200 V/m maintained). They
found the sludge to contain 98.29% maghemite, Fe2O3, and 0.26% angelellite, Fe4As2O11
plus some arsenic trioxide, As2O32. The same group also measured arsine gas, AsH3,
during electrolysis (see Section 2.5.3). Parga et al. (2005b) used XRD, scanning elec-
tron microscopy (SEM), and transmission Mossbauer spectroscopy (TEM) to find non-
stoichiometric magnetite particles, goethite, and amorphous iron oxyhydroxides in the
air-dried sludge produced from EC with air injection using carbon steel electrodes (cur-
rent density j = 3.7-4.6 mA/cm2, pH =7). Arienzo et al. (2002) characterized the
products from EC using XRD to find poorly ordered Fe mineral phases with a dom-
inance of 5-line ferrihydrite, and including no lines of ferric arsenate phases (j= 0.4
mA/cm2, pH 6.5). Gomes et al. (2007) used a combination of XRD, Fourier transform
infrared spectroscopy (FT-IR), X-ray photon spectroscopy (XPS), SEM, and Mossbauer2Pinisakul et al. (2002) noted that Fe(OH)3 could be converted to Fe2O3 during the ignition of sludge
samples at 700, which is a requirement of XRD analysis. They also reported at one point a sludgecomposition of 98.29% Fe2O3 and 0.26% As2O3 (replacing angelellite for arsenic trioxide). It is notknown which is correct.
80
spectroscopy to find well crystalline phases such as magnetite, and poorly crystalline
phases, such as iron oxyhydroxides and lepidocrocite, along with amorphous or ultrafine
particulates in the floc produced by EC with iron electrodes (j = 3 or 30 mA/cm2, pH
= 0.64)3.
2.6 Chapter Summary
A basic overview of arsenic in natural waters has been given. The predominant
forms of arsenic in natural waters (pH 6-8) include H2AsO –4 and HAsO 2 –
4 for As[V] and
H3AsO3 for As[III]. Numerous insoluble iron (hydr)oxides have been reviewed which can
be formed as iron is introduced to water. Iron oxide minerals tend to form in poorly
crystalline states with high specific surface area such as amorphous-Fe(OH)3 (sometimes
called Ferrihydrite) but, over time, undergo transformations to more crystalline forms
such as goethite or hematite.
Adsorption, or surface complexation, is the accumulation of a substance at or
near an interface relative to its concentration in the bulk solution. Sorption is a term
that encompasses adsorption, absorption, and surface precipitation, distinct from solid-
solution formation in which a stoichiometric new phase is formed. Adsorption of arsenic
onto ferrihydrite, goethite, and other iron (hydr)oxides has been studied extensively and
tends to follow a Langmuir type isotherm and consist of strong inner-sphere binuclear,
bidentate complexes. Iron (hydr)oxides tend to have a higher affinity for As[V] than
As[III] at the pH of natural waters, though the affinity order can change depending on3Both 3 and 30 mA/cm2 were used in reported batch experiments, but the paper does not state which
experiments were used in sludge product analysis
81
pH. Arsenate adsorption tends to be maximized near pH 4, while arsenite adsorption
tends to be maximized near pH 6-7. Sorption of arsenic in a lab setting is commonly
achieved by either coprecipitation or post-synthesis adsorption (PSA). In coprecipitation,
iron (hydr)oxides are formed in the presence of arsenic oxyanions, usually by the chemi-
cal addition of ferric salts. In PSA, iron (hydr)oxides are formed prior to the addition of
arsenic. Adsorption of arsenic onto zero-valent iron (ZVI) in solution is due to the con-
tinuous corrosion of the iron, producing Fe 2+ and/or Fe 3+ ions which quickly hydrolyze
to form iron (hydr)oxides adsorbent.
The kinetics of adsorption onto amorphous ferrihydrite, and many other oxides,
includes an initial rapid step, typically between 5-30 min, and is sometimes followed by
a slower step during which the adsorption density may increase at a rate of a few per-
cent per day. Coprecipitation experiments producing equivalent amounts of ferrihydrite
showed much higher adsorption rates at pH 7.5-9.0, with >95% As[V] adsorbed after 5
min and approximately twice as much As[V] adsorbed compared to PSA experiments
after 24 hours. However, arsenic can begin to desorb from the copreciptate as the co-
precipitate ages, approaching the capacity of PSA. The kinetics of arsenic removal via
ZVI depends on the rates of several subprocesses, including the corrosion rate of Fe(0)
(which depends on the DO concentration and pH), the rate of formation of ferric hy-
droxide precipitates, as well as the rate of arsenic adsorption onto the precipitates.
Many inorganic and organic aqueous species can affect arsenic adsorption. They
can compete with arsenic directly for available binding sites or indirectly by altering the
electrostatic charge at the adsorbent surface. Both processes are influenced by pH, ion
82
concentration, and the intrinsic binding affinity of the adsorbent. Phosphate, silicate,
and carbonate are known to decrease arsenic adsorption, with the combined effect that
is worse than the sum of individual effects. Calcium, magnesium, and sulfate have little
to no detrimental effect on arsenic adsorption.
A simple electrochemical cell consists of two electrical conductors (electrodes)
separated by an ionic conductor (electrolyte). Numerous processes and factors can affect
the transport of charge across the electrode/electrolyte interface. The cell potential
must be large enough to compensate for the desired current density (governed by the
Butler-Volmer relationship) as well as the ohmic drop across the electrolyte. To minimize
the ohmic drop and obtain an isolated measurement of the interface potential, a three-
electrode cell is often used.
Electrocoagulation (EC) is a complicated process in which consumable elec-
trodes supply ions that can be used in situ as coagulants or sorbates. The applied
electric field can also aid coagulation and aggregation of particles into flocs. EC using
iron can generate either Fe 2+ or Fe 3+ ions which may oxidize (in the case of Fe 2+) and
quickly hydrolyze to form insoluble iron (hydr)oxides. The exact composition of iron
(hydr)oxides end products is a sensitive function of potential, pH, and concentration of
various species in solution, making it difficult to predict. However, the potential can be
used to control the composition. If the process is galvonostatic (constant current), then
the current density j can be used to control the potential within a small range via the
Butler-Volmer relationship. Other controlling parameters include the charge density q,
which measures the number of iron ions dissolved into solution and hence the dosage of
83
iron (hydr)oxides, and the current processing time tCP which controls the complexation
coagulation time. The parameters j, q, and tCP can only be adjusted independently of
one another if the electrode surface area to active volume ratio, A/V , is adjusted as a
compensatory variable.
A major advantage of EC is the ability to oxidize As[III] ions to As[V] ions,
which more readily form complexes with iron (hydr)oxides at neutral pH. The mechanism
of arsenic complexation in EC is likely adsorption of As[III] and As[V] oxyanions onto
iron (hydr)oxides, with some As[III] oxidizing to As[V] before adsorption.
EC has been studied in bench scale reactors to remove arsenic from wastewater
and potable water. Iron electrodes have been found to be more efficient that aluminum
or titanium electrodes. Few studies stress or monitor any additional complexation time
following the electrolysis step in EC, though there is potential to increase efficiency by
separating the iron (hydr)oxides dose (set by current processing time) from the com-
plexation time. Many studies report that initial arsenic removal is rapid, followed by
a slower phase, though none fit data to a kinetic model. A first order model was fit
for fluoride removal with aluminum electrodes, but this model differs from arsenic re-
moval in that there is no analogue to the oxidation of As[III] and possibly Fe 2+. Several
studies have characterized the sludge containing corrosion products produced via EC.
Amorphous iron oxyhydroxides, including ferrihydrite, are often found, along with some
smaller more crystalline phases.
84
Chapter 3
Chemical and Physical Analysis
of Arsenic Complexation with
Iron in ECAR
3.1 Introduction
3.1.1 Operating parameters in ECAR
It is clear from Chapters 1 and 2 that the parameters in ECAR - current
density, j, charge density, q, current processing time, tCP , mixing time, tM , and filtration
time, tfilt - may have significant and potentially competing influences on performance
considerations relevant to rural Bangladesh - arsenic removal effectiveness, treatment
time, and operating costs. Chapter 2 described how these influences depend on specific
properties of the water being treated, such as pH and the composition of co-occuring
85
solutes. To better understand the influences of ECAR parameters on performance criteria
in rural Bangladesh, empirical studies must be performed in the specific groundwater
environment of Bangladesh. Exploratory studies to understand the cause of parameter
influence are also essential to better predict and optimize ECAR performance, specifically
with regard to lowering costs (a key consideration for rural Bangladesh).
The parameters of ECAR - j, q, tCP , tM , and tfilt - and the processes they
govern are described briefly in Section 1.4.1 and in more detail in Section 2.5.2. It will
be assumed that the reader is familiar with these parameters and their potential effects.
To study the influence of parameters on performance, it is important to define
measurable performance metrics. Performance goals relevant to rural Bangladesh include
effective arsenic removal to levels below the legal Bangladesh limit (50 µg/L) or the
WHO limit (10 µg/L), a treatment time within about 6-7 hours (for community scale
treatment; see Section 1.4.2), and operating costs that are low enough for clean water to
be affordable to those making less than US $1 per day (for more discussion see Chapter 1).
Effective removal for Bangladesh is demonstrated by reducing relevant levels of arsenic
in a relevant water matrix to less than the legal Bangladesh or the WHO limit. For
this work, arsenic removal capability, or just removal capability, will refer to the ability
to reach a predefined low arsenic concentration (usually the Bangladesh or WHO limit)
given a specific initial arsenic concentration and water matrix (i.e. synthetic or real
Bangladesh groundwater). The metric for treatment time will simply be the treatment
time, referring to the total time for ECAR treatment, ttreat = tCP + tM + tfilt, where
tCP is the current processing time (i.e. the time of current flow through the water
86
- see Equation 2.34), tM is the mixing time, and tfilt is the total time of filtration. A
measurable metric for operating cost is more subtle. The term arsenic removal efficiency
is often used in literature a metric of dosage and cost, but the definition varies and is not
always explicitly given. As discussed in Section 1.4.2, both iron and energy consumption
(which determine operating costs) depend on the charge required to reduce arsenic from
the initial to the final acceptable level. To act as a metric of operating costs, arsenic
removal capacity, or just removal capacity, will be used to refer to the average arsenic
removed per coulomb to reach a specified arsenic limit (usually the Bangladesh or WHO
limit):
Arsenic Removal Capacity =[As]i − [As]limit
qlimit(3.1)
where qlimit is the charge density required to reduce arsenic from initial concentration
[As]i to the limit concentration [As]limit. Operating costs are inversely related to removal
capacity (i.e. higher removal capacity will lead to lower costs), though it should be
noted that this does not take into account the effect of operating voltage on energy
consumption.
The influence of ECAR parameters on performance metrics and the cause of
that influence depends on a broad and diverse set of variables covering multiple dis-
ciplines. It is necessary to narrow the scope to understanding processes and variables
that are most relevant to rural Bangladesh and that could potentially be manipulated
to improve performance. In addition, although electrocoagulation is not new, very few
studies exist to understand the underlying process of arsenic removal (see discussion in
Section 2.5.5). Thus broad exploratory techniques are useful to help direct more specific
87
research in the most relevant directions.
Batch tests are a simple yet effective tool to understand the influence of various
parameter sets on arsenic removal capability, treatment time, and removal capacity in
the specific groundwater environment of Bangladesh. This common technique involves
treating small batches of water in the same initial state with some specific combination
of operating parameters and comparing outcomes. Batch tests are limited by the ability
to recreate the same initial water state and experimental conditions, though variability
can be measured with repeat experiments. They can be designed to isolate the effect of
single variables or explore the effect of different relevant variable sets, narrowing down
the most relevant variables for further study.
Voltammetry is a sensitive technique used to probe the electrochemical behavior
of a system. It involves adjusting the potential of an electrochemical cell and observing
the resulting current response or polarization curve (i.e. an i-E curve, discussed in
Section 2.5.1). Slow scan rates can be used to generate polarization curves that correlate
a specific current density to the electrode interface potential in the solution of interest
(using a three-electrode cell; described in Section 2.5.1). This scan can be compared
to thermodynamic equilibrium potentials to identify the likely electrochemical reactions
occurring at the iron anode at different current densities in ECAR. Polarization curves
can also be used to gain important information about background processes that may
affect energy consumption or arsenic removal capacity by decreasing the current efficiency
of iron dissolution (discussed in Section 2.5.2).
Electrochemical impedance spectroscopy (EIS) is a powerful electrochemical
88
technique that uses the measured impedance of a system to infer information about
underlying reaction pathways or to separate and measure electrical properties of the sys-
tem, such as electrolyte resistance. EIS provides a more accurate measure of electrolyte
resistance than using a conductivity probe because electrolyte resistance depends on
the geometry of the area in which the current is carried (to convert conductivity probe
measurements to electrical resistance, one must assume that there is a uniform current
distribution through a definite electrolyte area - rarely true in an electrochemical cell).
Electrolyte resistance is a factor in the energy consumption, and hence operating costs,
of ECAR treatment.
3.1.2 Research Objectives
The primary goals of this research were to demonstrate the arsenic removal
capability of ECAR in the groundwater environment of Bangladesh, and to explore
the influences of ECAR parameters on arsenic removal capacity for the purposes of
optimizing costs within removal capability and treatment time constraints. Based on
previous electrocoagulation research and expectations from electrochemical theory, initial
focus was placed on parameters j and q with some exploration of tM .
We began by designing a bench-top electrochemical cell and reproducible proto-
col for ECAR treatment. Two recipes were developed to generate reproducible batches
of synthetic Bangladesh groundwater (SBGW) - one version with low phosphate and
silicate and one version with average levels of phosphate and silicate. Batch tests were
used to look at the effect of j and q on arsenic removal capacity and treatment time in
89
the ranges j = 0.02 - 100 mA/cm2 and q = 0 - 175 C/L. Additional batch tests were used
to explore the influence of tCP and tM . Polarization scans, batch tests, and adsorption
experiments with post-synthesis ECAR-generated iron (hydr)oxides adsorbent were used
to explore the cause of current density trends in removal capacity, study adsorbent aging
and affinity for arsenic, and understand the mechanism of As[III] removal. Information
gained was used to suggest initial parameters for ECAR field tests and identify regions
of parameter space for further study.
Specific Research Objectives:
1. Design a laboratory method for ECAR, keeping in mind eventual scale-up.
2. Demonstrate the capability of ECAR to reduce high levels of arsenic (>98% sur-
veyed wells) to below the WHO recommended maximum limit of 10 µg/L in syn-
thetic Bangladesh groundwater constituted to mimic key properties relevant to
arsenic removal.
3. Characterize arsenic removal capacity of ECAR with respect to current density, j,
and charge density, q, in synthetic Bangladesh groundwater.
4. Explore the cause of current density trends in arsenic removal capacity and the
mechanism of As[III] removal.
5. Identify trends and tradeoffs in removal capacity and treatment time between op-
erating parameters.
6. Identify an optimal range of current density and charge density for Bangladesh,
90
taking into account arsenic removal effectiveness, treatment time, and minimization
of operating costs.
7. Explore low cost alternatives to membrane filtration in ECAR.
3.2 Methods
3.2.1 Arsenic analysis and arsenic speciation
Arsenic in solution was measured using Inductively Coupled Plasma Mass Spec-
troscopy (ICPMS) analysis provided by Curtis & Tompkins, Ltd (Berkeley, CA). ICPMS
combines an inductively coupled plasma as a method of producing ions with a mass spec-
trometer as a method of separating and detecting the ions. It is one of the most accurate
available methods for trace detection of metals. Curtis & Tompkins (C&T) follow EPA
procedure 6020 (EPA, 1998) with a reporting limit for arsenic of 1 ppb and reported ac-
curacy of 10%. Repeated standard samples sent to C&T suggest a precision of <4%. All
samples were digested prior to being sent to C&T using 1 mL of 10% HNO3 (Mallinck-
rodt) per 9 mL sample. All results received from C&T were multiplied by a factor of
10/9 to compensate for the dilution. One sample containing only deionized water and
HNO3 was sent with each group of samples to measure the current arsenic level in the
HNO3 (usually <1 ppb, or below detection limit) to ensure no arsenic contamination
from the preservative and to provide a check on C&T procedures.
Sending samples to C&T and obtaining results for ICP-MS typically took about
10 days. Immediate (< 10 min) arsenic measurements for internal purposes and some
91
field measurements were often performed using a Rapid Arsenic Quick Test (Industrial
Test Systems, Inc, SC). The Quick Test is a colorimetric test capable of detecting As[V]
and As[III] based on the reduction of arsenic compounds to arsine gas. Reported accuracy
is 8-80% (Abbgy et al., 2002). In general, we found Quick Test measurements to be
within 20% of ICPMS measurements over a range of 0 - 500 ppb (all samples >30 ppb
were diluted to <30 ppb for testing). Reported errors on Quick Test measurements are
based on the separation between adjacent colorimetric levels on the company-provided
color chart (5 - 20 ppb depending on concentration). Most Quick test results were later
verified via ICPMS. Thus all arsenic measurements reported here are based on ICPMS
unless otherwise stated.
As[III] and As[V] were separated in water using disposable ion exchange columns
(Metalsoft Center, Highland Park, NJ) containing 2.5 g of aluminosilicate adsorbent ca-
pable of retaining As[V] but not As[III] in the pH range 4-9 (Meng & Wang, 1998). Per
manufacturer’s instructions, only 50 mL of effluent was accepted from each cartridge
with the first 5 mL discarded. The cartridges were rated for a maximum total arsenic
concentration of 500 µg/L, so samples were first diluted with deionized water for con-
centrations above this range. The flow rate was maintained at < 90 mL per minute
using a plastic syringe with a luer tip. Meng et al. (2001) found that the cartridges
removed 98% of the As[V] and less than 5% of the As[III] in spike and recovery tests
within the specified operating range. Lab tests of deionized water spiked with 500 and
400 µg/L As[V] (as Na2HAsO4 ∗ 7H20) showed 96% and 100% removal respectively (via
Quick Test), and similar tests with 100 µg/L As[III] (as NaAsO2) showed no arsenic re-
92
moval, consistent with results of Meng & Wang (1998). The cartridges require complete
filtration of particulate arsenic before use, and so could not be relied upon to accurately
measure As[III] in solutions in which rapid precipitation of iron (hydr)oxides was taking
place (i.e. within 20 minutes of adding iron to arsenic-layden water at neutral pH).
3.2.2 pH, dissolved oxygen measurements
All pH measurements in the lab and in the field were performed using a Sym-
pHony Ag/AgCl sealed gel electrode with a built in thermistor for automatic temperature
compensation connected to a VWR SympHony Meter (model SP8OPD). The electrode
was permanently filled with a KCL gel and was operated and maintained according to
factory instructions. The probe was 2-point calibrated every two days of operation using
a pH 7.01 buffer (Orion) and one of either a pH 4.01 or pH 10.01 buffer (Orion).
Dissolved oxygen measurements were performed using a SympHony Dissolved
Oxygen Probe containing a polargraphic sensor connected to the SympHony meter de-
scribed above. The probe was operated and maintained according to factory instructions.
Before each daily use, the probe was calibrated in water-saturated air (air calibration
was recommended by the manufacturer for both air and aqueous measurements). A
second 0-oxygen calibration point is recommended, but not required, for measurements
of < 2 mg/L, however this was not routinely performed since the majority of measure-
ments were above this range. Note however that this may slightly increase the error on
measurements < 2 mg/L reported below.
To determine probe errors in a solution of similar ionic strength to synthetic
93
Bangladesh groundwater (SBGW), 20 consecutive pH and DO measurements were taken
of leftover SBGW (recipe SBGW-1; described fully in Section 3.2.3 below) which was
several days old and had reached pH equilibrium (i.e. the pH no longer drifted up). The
pH probe was freshly rinsed in 0.1 M HNO3 followed by deionized (DI) water. Both
probes were rinsed in DI between each measurement. Average pH was found to be 9.33
±0.06, and average DO was found to be 2.77 ±0.18 mg/L with no significant outliers.
The values 0.06 and 0.18 mg/L were adopted as the measurement error on the pH and
DO probe respectively.
3.2.3 Development of Synthetic Bangladesh Groundwater (SBGW)
Synthetic Bangladesh groundwater (SBGW) is essential for standardized repro-
ducible work. Preparing and storing SBGW requires attention and care. This section
describes these steps in detail to aid future experimenters.
Choosing water parameters
The country of Bangladesh is approximately the size of Iowa and contains sev-
eral distinct geological regions with potentially very distinct groundwater chemistry.
Arsenic occurs in the groundwater beneath the Halocene floodplains of the Ganges,
Brahmaputra and Meghna Rivers in the Bengal Basin (Davies, 1995; Ravenscroft, 2003).
Hand-pumped tube wells access groundwater 20-75 m deep (Burgess & Ahmed, 2006)
where lateral flow is very low, creating pockets of relatively similar aquifer chemistry in
regions where high arsenic is found. Groundwater of the Bengal Basin is generally of the
94
Ca2+ (Mg2+)−HCO−3 type, with a near-neutral pH, a relatively high level of mineral-
ization (the conductivity is generally 500-1000 µS/cm) and HCO –3 commonly present
at several hundred mg/L (Burgess & Ahmed, 2006). The water is predominantly anoxic
and strongly reducing (Burgess & Ahmed, 2006), though oxygen levels can increase by
several mg/L as the water is pumped from the tube well and into the open air.
The specific groundwater properties that are key to mimicking ECAR perfor-
mance in Bangladesh are those which can affect the rate of iron dissolution or those
which affect the sorption of arsenic onto iron (hydr)oxides. The pH, water conductivity,
presence or absence of oxygen, and ions in high concentrations that are electrochemi-
cally active in our operating region could affect iron dissolution. The pH, oxygen level
(through its effect on arsenic speciation), and competing co-occuring solutes could affect
arsenic sorption onto iron (hydr)oxides.
To prepare synthetic groundwater with ion concentrations mimicking conditions
in Bangladesh, we initially followed the composition developed by Petrusevski et al.
(2002) for background ions, with exception of Fe and Mn. Fe was left out because the
small amount present (0.03 mg/L) would be overwhelmed by the iron from our sacrificial
anode. This first recipe (SBGW-1) is presented in Table 3.1. It is characterized by having
low phosphate, no silicate, and high sulfate (to make up for ionic strength).
To understand the importance of interfering ions such as phosphate and silicate
(see Section 2.4.6) on arsenic removal, a second water batch was developed. To determine
representative values of key water properties for the second batch, the British Geological
Survey (BGS) was used (BGS, 2001). The BGS, conducted 1998 - 1999, consists of
95
Table 3.1: Synthetic groundwater composition and corresponding mean values and stan-dard deviations of groundwater in Bangladesh, obtained from the BGS database (valuesfor pH, As[III], As[III]/Astot, HCO –
3 , and Cl – taken from special Study Areas; all othervalues were taken from National Survey data, using only wells with As > 10 µg/L fordata analysis, and setting values below detection limits to zero). For each solute, thepercentage of wells in the BGS database with a concentration (x) less than that inthe synthetic groundwater 1 and 2 (xs1, xs2) is given. Values for pH, As, As[III], andAs[III]/Astot are measured values across water batches, all other ions are spiked values.
SBGW 1 % of BGS wells SBGW 2 % of BGS wells BangladeshLow P, no Si with x <xs1 Ave P, Ave Si with x <xs2 Groundwater (BGS)
pHinitial 7.04 ±0.42 50a 7.06 ±0.16 47a 7.05 ±0.22As (µg/L) 581 ±75 98 556 ±29 98 129 ±155As[III] (µg/L) 304 ±41b 92a 288 ±19b 92a 91 ±136As[III]/Astot 0.52 ±0.05b 52a 0.55 ±0.02b 52a 0.46 ±0.41HCO –
3 (mg/L) 239 0a 275 6a 501 ±144P (mg/L) 0.01 5 1.3 67 1.3 ±1.5Si (mg/L) 0 0 19.5 51 19.7 ±5.1
SO2−4 (mg/L) 81 99 8 88 4.6 ±17.4
Ca (mg/L) 53 52 61 56 66 ±53Mg (mg/L) 8 16 8 16 27 ±21Cl (mg/L) 125 87a 125 87a 80.8 ±203Na (mg/L) 134 82 138 82 94 ±183
a Comparison to Special Survey Areas of BGS only (155 tube wells over a small region)b Data available for 6 water batches only (As[III] was not routinely measured)
96
samples from 3534 tubewells from 61 of the 64 districts of Bangladesh and from 433 of
the 496 Upazilas. The samples covered approximately 0.03 - 0.05% of all tubewells in
Bangladesh and is believed to be representative of all wells. In addition, three special
study areas were established in the Upazilas of Nawabnanj, Faridpur, and Lakshmipur
in which pH, dissolved oxygen, and a wide range of additional trace elements were mon-
itored. These study areas are not considered representative of all wells in Bangladesh,
and indeed contain a greater proportion of deep wells than found in Bangladesh as a
whole, so statistics from these areas were treated with caution.
To determine the average concentration of trace elements in arsenic-contaminated
tube wells, only wells with [As] ≥ 10 µg/L were used (1485 wells in the full survey, 155 in
the special surveys). Relevant ions (aside from arsenic) were chosen as those with aver-
age concentrations > 0.3 mg/L in the full survey (excluding K, Mn and Fe). Exceptions
include relevant ions only available in the special survey, such as Cl – (known to enhance
rusting) and HCO –3 (known to compete with arsenic). Average values of relevant ions
from the BGS are listed in Table 3.1. In calculating averages, values below the detection
limit were set to zero (detection limits and percentage of wells below detection limits are
also presented in Table 3.1).
The special Survey Area results can be augmented by smaller surveys over wider
and more representative geographic areas. Bhattacharya et al. (2002) surveyed 32 wells
from 9 affected districts over a much larger geographical range than the special survey
areas of the BGS, covering the full extent of the Ganges, Brahmaputra and Meghna
rivers in Bangladesh. They measured HCO –3 concentrations of 320-600 mg/L (though
97
22% of wells had HCO –3 < 300 mg/L), Cl – of 1.9-794 mg/L, and As[III]/Astot of 67-99%.
Electrical conductivity varied from 307-1970 µS/cm. These values are consistent with
those in the BGS, though they suggest that HCO –3 < 300 mg/L is not as rare as the
BGS suggests and that As[III] may be more prevalent in general than As[V].
The total arsenic concentration chosen for synthetic water was 600 µg/L, greater
than that found in 98% of wells surveyed by the BGS (actual measured values tended to
be a bit lower, represented in Table 3.1). Of the 600 µg/L, half was inserted as arsenite
(As[III]) and half as arsenate (As[V]) to reflect the fact that both appear together in
Bangladesh as well as take into account the difficulty of removing As[III] relative to
As[V].
The initial pH of all synthetic batches was set to 7.0 ±0.4 to reflect the near-
neutral pH of Bangladesh groundwater. The dissolved oxygen (DO) content of tube well
water reported in the BGS (from a special Study Area and wells with As > 10 µg/L) is
0.16±0.69. However the act of pumping the water from the aquifer to the surface and into
a bucket raised the oxygen level by 2-3 mg/L. To reflect this, dissolved oxygen content
in synthetic water was set to < 3.00 mg/L, though in some select cases experiments
were done with up to 4.00 mg/L. Average DO levels for SBGW-1 and SBGW-2 were
2.27 ±0.50 mg/L and 2.91 ±0.40 mg/L respectively. In terms of the presence or absence
of oxygen (which may be more important than the actual concentration), oxygen was
present for these experiments. Equilibrium DO concentrations in the atmosphere were
measured to be 7-9 mg/L.
98
A reproducible protocol for synthetic water preparation
To create synthetic Bangladesh groundwater, stock solutions of arsenate (As[V])
and arsenite (As[III]) were prepared from Na2HAsO4·7 H2O (Sigma) and NaAsO2 (Ricca)
fully dissolved in deionized water. Deionized water used for stock As[III] solutions was
previously purged with N2 and tightly capped to prevent As[III] oxidation during storage.
The ratio of As[III] to Astot was monitored every two weeks. This allowed us to ensure
minimal conversion of As[III] to As[V] - if the ratio was less than .95, the solution
was discarded and remade. Phosphate stock solutions were made from Na2HPO4 ·7 H2O
(EM Sergeant). Bicarbonate stock solutions were made from NaHCO3 (Baker). Calcium
Sulfate stock solutions were made from CaSO4 ·2 H2O (MP Biomedicals). A mixed stock
solution was made from dissolving appropriate amounts of MgCl2 ·6 H2O (Mallinckrodt),
NaCl (Mallinckrodt), and CaCl2 (Merck) in DI water. All chemicals were reagent grade.
To begin a water batch (typically 8 liters), stock solutions (excluding As[III])
were mixed together according to the appropriate recipe for SBGW-1 or SBGW-2 (see
Table 3.1) and purged with N2 for approximately 1 hour to reduce the dissolved oxygen
content, leaving a clear solution with a pH of about 8 and DO < 2.0 mg/L. The pH was
then lowered by bubbling excess CO2 gas (at 3 psi) into the water for 1 min or 4 min,
leaving a clear solution at about pH 7.0 or pH 6.0 for SBGW-1 or SBGW-2 respectively.
For SBGW-2, silicate was added from a daily prepared stock solution (Na2SiO3·5H2O
dissolved in 200 mL deionized water). To prevent any polymerization of the silicate, the
alkaline stock solution was rapidly mixed into the slightly acidic synthetic groundwater
(pH 5.8-6.4 with excess dissolved CO2) where it was allowed to equilibrate for at least
99
1 hour, or in some cases overnight. For SBGW-2 the pH was then raised to 7.0 ±0.4
by bubbling compressed N2 into the solution (in some cases this was not necessary after
the solution sat overnight). Finally, As[III] stock solution was added, resulting in a
final total arsenic concentration of 600 µg/L. The synthetic groundwater was stored in a
loosely covered plastic carboy with N2 continuously supplying the headspace to maintain
a low DO level. The DO of synthetic groundwater in the storage container tended to
be < 1 mg/L (lowered during rapid bubbling of CO2), however decanting the synthetic
groundwater for testing and pH measurement generally caused the DO to rise to 2-3
mg/L. Water batches were stored for a maximum of 2 days until the pH of the solution
rose outside of the required range (pH 6.8-7.2). In some cases, As[III] and As[V] stock
solutions were added to synthetic groundwater immediately before testing and stirred
for 2-3 minutes before taking initial pre-treatment samples.
Water batch behavior
While the Astot concentration and dissolved oxygen concentration in synthetic
groundwater batches remained stable over several days, the pH and As[III] concentration
showed change. The pH tended to drift up by about 0.3 pH units per day, quickly
leaving the required range of pH 6.8-7.2. The As[III] concentration tended to decrease
rapidly as well, demonstrated by the ratio of As[III]/Astot over 2-3 days for several water
batches of the SBGW-1 type (Figure 3.1). Typical behavior, demonstrated by water
batch 38 in Figure 3.1, shows almost complete oxidation of As[III] to As[V] after only 3
days. This is much faster than rate of As[III] oxidation due to oxygen alone, which is
100
known to be kinetically slow. Cherry et al. (1979) found only 25% oxidation of As[III]
after 6 months in water saturated with pure O2 at pH 7 compared to no oxidation in
deoxygenated water. Eary & Schramke (1990) cite a half-life of 1-3 years for As[III] in
water equilibrated with atmospheric oxygen. Complete or near complete oxidation of
As[III] within 2 days was observed by Devitre et al. (1991) in the presence of synthetic Fe
oxyhydroxides and Mn oxyhydroxides, however neither Fe or Mn are present in SBGW-1
batch before electrochemical treatment. Other known oxidizing agents, such as chlorine,
potassium, permanganate, peroxide, ozone, and light are either not present in SBGW
or unable to completely oxidize As[III] on this rapid timescale. No significant As[III]
oxidation was seen in deionized water spiked to 300 µg/L of As[III] and 300 µg/L As[V]
(similar to SBGW) after 8 days of monitoring in the same container used to make SBGW,
either in equilibrium with the atmosphere or with a DO < 2.0 mg/L (maintained through
constant N2 purging of the headspace). 75% oxidation after 5 days was observed (using
Quick Test results only) using the SBGW-1 recipe in equilibrium with the atmosphere,
with and without excessive bubbling of CO2.
As[III] is known to form complexes with calcite (Roman-Ross et al., 2006).
Calcite is oversaturated in SBGW, and its formation would lead to an increase in the
pH over time, as observed. It is possible that calcite could capture some As[III] and
settle to the bottom of the water batch storage container, effectively reducing the As[III]
concentration. However, this would lead to a decreasing Astot concentration over time,
and Astot was observed to be stable. Careful studies to discover the cause of the rapid
oxidation of As[III] were not pursued in favor of simply cutting off the useful lifetime of
101
each water batch at 1 day.
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.0
0.2
0.4
Water Batch Age (days)
AsI
II/A
s−to
tal
WB 38WB 43WB 45
Figure 3.1: The ratio of As[III] to total-arsenic (Astot) over a period of 2-3 days forthree representative water batches (WB) of the SBGW-1 type. For water batches 38and 45, no initial (0 day) measurement of As[III] was done, so average As[III]/Astot ratioof monitored water batches (8 total) at day-0 was used with the standard deviation asan error.
3.2.4 Electrochemical cell
Two benchtop electrochemical cells were used in experiments, each consisting
of a glass beaker containing a pure iron wire (diameter = 0.18 cm) anode coiled into a
spiral and a copper mesh cathode. The active area of both iron and copper electrodes
varied with experiment, ranging from 9-150 cm2 for the iron and 24 - 150 cm2 for the
copper mesh. The iron wire was tested for iron content by Curtis & Tompkins Lab
(Berkeley, CA) using EPA preparation procedure 3050B and testing procedure 6010B
and found to contain 106 mg/kg iron (essentially 100% iron). Both cells were hooked
up in series with an EG&G model 173 Potentiostat/Galvonostat in galvonostatic mode
102
with either a model 376 logarithmic current converter or model 176 current follower (all
from Princeton Applied Research). This was in series with a Kiethley model 173A or
Fluke model 73111 multimeter set to measure current. The setup for each cell is shown
in Figures 3.2 and 3.3.
The first of the two cell configurations was a 850 mL-separated cathode cell
shown schematically in Figure 3.2. The 1 liter cell beaker had a diameter of 11 cm. The
iron electrode, coiled to a diameter of about 10 cm, was submerged to approximately
4 cm above the beaker bottom to make room for a magnetic teflon-coated stir bar. A
Schott porous glass frit filter, consisting of a glass beaker (diameter = 4.3 cm) with
a porous glass frit bottom (allowing charge, but very little water to pass), was placed
approximately 2 cm above the anode with the copper cathode placed inside the glass
beaker, creating a separated cathode chamber. The top of the glass filter was held above
water level to ensure minimal mixing of the cathodic and anodic solutions. Sample
water filled both chambers up to a total volume of 850 mL (approximately 40 mL in the
cathode chamber). Separation of the cathode was intended to shield the main treatment
volume from the reduction of As[V] to As[III] at the cathode and the pH increase caused
by H + reduction to H2 gas. Minimum separation between the cathode and anode was
approximately 2.5 cm.
The second cell configuration was a 3 L-single chamber cell shown schematically
in Figure 3.3. The cell body was a 3.5 L glass beaker with a diameter of 15.3 cm. A flat
spiral iron electrode (same shape and material used in the 850 mL-separated cathode cell)
was submerged 1/6 of the total water height (13 cm from the bottom), with the square
103
copper mesh electrode placed flat on top. Two layers of refurbished polyvinylidene
fluoride hydrophilic membranes with 0.1µm pore size (Millipore; see Section 3.2.5 for
refurbishing procedure) were placed between the two electrodes for insulation, leading
to an electrode separation of 0.05 - 0.5 cm (range varied as the iron electrode coil was
not completely flat).
Galvonostat
Counter Lead Working Lead
+ -
Ammeter
Stir Bar
Iron Anode
Copper Cathode
Porous Glass
Frit
Figure 3.2: Schematic representation of 850 mL-separated cathode electrochemical cell.
104
Galvonostat
Counter Lead Working Lead
+ -
Ammeter
Stir Bar
Iron Anode
Copper Cathode Insulating
Membrane
Figure 3.3: Schematic representation of 3 L-single chamber electrochemical cell.
105
3.2.5 Electrochemical Arsenic Remediation Procedure
ECAR treatment consists of three stages in series; (1) electrochemical dosing,
(2) mixing, and (3) filtration. The dosing stage is characterized by electrodes in contact
with the contaminated water, as a galvonostatic (constant) current I is passed using
an external voltage source. During this stage, iron dissolves in amounts determined by
Equation 2.19, forming EGA sludge (i.e. iron (hydr)oxides) in solution. Other processes
that may occur during the dosing stage are discussed in Section 2.5.2. During the
mixing stage, external voltage is off and electrodes are removed from the water, while
the dosed solution is stirred or lightly agitated to facilitate further complexation between
the arsenic in the water and EGA as well as coagulation of EGA and/or arsenic-EGA
complexes to allow for filtration. In the filtration stage, arsenic-laden EGA is separated
from the water using some kind of filtration process.
The basic methods used for each of the three ECAR stages during all lab tests
are described below. Specific experiments are fully described by adjustable ECAR oper-
ating parameters q, j, tCP and tM (tfilt is generally determined by the filtration process)
along with experimental conditions, including the electrochemical cell parameters (des-
ignated by A/V and the cell design), and the test solution composition (generally des-
ignated by SBGW-1 or SBGW-2). Experiment-specific procedures and/or experiment-
specific modifications to the basic method are described in the subsequent sections.
106
Dosing method
The dosing stage began by filling the empty electrochemical cell (i.e. without
electrodes) to the appropriate volume (850 mL or 3 liters for the 850 mL-separated
cathode cell or the 3 L-single chamber cell respectively) with sample water (usually
synthetic Bangladesh groundwater). In some cases, the sample water was stored without
arsenic. In such cases, one or both of the As[III] and/or As[V] stock solutions were poured
into the empty cell beaker first, with the sample water poured on top to facilitate mixing.
Initial pH and DO measurements were taken in the beaker during mild stirring (via
a magnetic stirrer); also facilitating the mixing of As[III] and As[V] into the solution.
Initial pH and DO measurements took approximately 2-3 min, during which time the DO
tended to rise from <1.5 mg/L to 2-3 mg/L. If required, aliquots of solution to measure
pre-treatment Astot and As[III] concentrations were removed from the top using a glass
luer tip syringe. If the pre-treatment As[III] concentration was measured, approximately
half of the aliquot was immediately filtered through a speciation cartridge.
The electrodes were then inserted and connected according to Figures 3.2 or
3.3. The galvonostat was turned on at t = 0 (measured by a portable timer) to provide
constant current I, along with the magnetic stirrer using a low setting (just high enough
to create a visible vortex at the top of the cell). Current was monitored using either the
Keithley or Fluke digital multimeter (see Section 3.2.4. After a total current processing
time tCP , current and stirring were turned off and the electrodes were immediately
removed, thus completing the dosing stage. Dosed solution was immediately passed into
the mixing stage (within 1 minute or less).
107
Mixing method
The mixing stage consisted of stirring the dosed solution in an open beaker
for a fixed amount of time tM using a magnetic stirrer. A post-dosage pH and DO
measurement were often taken during the first 2-3 minutes of mixing. Solution would
pass from the mixing to the filtration stage within 5 minutes (sometimes waiting for an
available filter apparatus).
Filtration method
Mixed solutions were filtered using a pyrex flask vacuum filter (Kontes) and
90 mm diameter polyvinylidene fluoride hydrophilic membranes with 0.1µm pore size
(Millipore). Filtration of 150 mL took approximately 4-5 minutes.
Repeated filtration of the filtrate showed no change in arsenic concentration,
indicating that no water was leaking around the membrane and all particles > 0.1µm
were filtered out in the first pass. Results of filtration with pore sizes up to 0.8µm
(discussed in Section 4.3.1) showed no increase in arsenic concentration, also indicating
full removal of particles > 0.1µm.
Membranes were refurbished one time after use in batch tests by soaking in
dilute HCl and rinsing twice in deionized water. Filtration of mixed water using re-
furbished membranes yielded the same arsenic concentration (within 5%) as filtration
with new membranes. Refurbished membranes were used to insulate the anode from the
cathode in the 3 L-single chamber cell only (see Section 3.2.4) - they were not used for
filtration during batch tests.
108
Lab practices
Before treatment, the electrochemical cell beaker was always rinsed in diluted
HCl followed by DI water to remove any buildup of iron (hydr)oxides precipitate or
arsenic from previous batch tests. All beakers used for aliquot transport were thoroughly
rinsed in DI water. The vacuum filter apparatus was doubly rinsed in DI water between
each filtration along with periodic rinses in dilute HCl. Sample vials for ICPMS testing
were unused and radiation sterilized by the vendor (BD Biosciences). Sterile gloves were
worn for all procedures.
3.2.6 ECAR batch experiments run 1 - Matrix 1
Purpose and experimental parameters
ECAR batch experiments were performed to (1) determine the minimum dose
required to reduce arsenic concentrations from 600 µg/L to below the Bangladeshi and/or
WHO maximum allowable concentration (50 µg/L and 10 µg/L respectively ) as well
as (2) study the effect of current density on arsenic removal capacity. In this case,
arsenic removal capacity refers to the amount of arsenic removed per coulomb passed (in
µg/L-Asremoved/C). This round of testing, known as Matrix 1 , used SBGW-1 synthetic
groundwater and the 850 mL-separated cathode electrochemical cell. Sixteen individual
batch tests were performed to cover all combinations of current density j = 0.07, 0.3,
0.7, and 1.1 mA/cm2 combined with current processing times tCP = 3, 5, 11, and 50 min
(total dose 1.4 - 390 C/L). All tests used one of two flat spiral iron anodes with an active
(submerged) area of 100 cm2, leading to a surface-area-to-volume ratio of 0.118 cm−1.
109
Each electrode was either sanded with fine sandpaper or rinsed in dilute HCl followed
by deionized water every three batch tests or every day, whichever came first. Initial
arsenic was measured within 1 hour of creating each synthetic water batch. In one case,
j = 1.1 mA/cm2 and tCP = 50 min, the initial arsenic was not measured, and the initial
arsenic for that test was taken to be the average of all water batches created for this run.
Establishing the mixing time
In order to set the appropriate mixing time, tM , for batch experiments, some
initial mixing kinetics tests were performed. In each test, dosed water was mixed for
some time with aliquots removed, filtered, and tested for arsenic using the Quick Test
(described in Section 3.2.1) every 5 minutes for 25 - 30 minutes. Results for two such
tests (kinetics test 1 and 2) after dosing at j = 1.1 mA/cm2, q = 39 C/L, tCP = 5 min are
shown in Figure 3.4. In both cases, the arsenic concentration falls rapidly for the first 10-
15 minutes of mixing, and much more slowly after that. In the case of kinetics test 1, the
concentration seems to stabilize after tM= 15 min. In kinetics test 2, the concentration
drops rapidly for about 5-10 minutes followed by a slow decrease continuing after 25
minutes.
Adsorption of As[V] onto pre-formed amorphous ferrihydrite at pH 8 is reported
to have an initial rapid uptake (< 5 min) followed by a slow uptake for as long as 8 days
(Fuller et al., 1993). Coprecipitation experiments producing the same equivalent amount
of ferrihydrite have shown rapid uptake (< 5 min) followed by slow uptake for about 24
hours, followed by very slow desorption continuing after 31 days of reaction (Fuller et al.,
110
0 5 10 15 20 25 30
5010
015
020
025
030
0
Mixing Time (min)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
9590
8580
7570
% A
rsen
ic R
emov
ed
Kinetics Test 1Kinetics Test 2
Figure 3.4: Fraction of remaining arsenic as a function of mixing time for SBGW-1.Both kinetics tests were performed under identical conditions after a dosing at j =1.1 mA/cm2, q = 39 C/L, tCP = 5 min, with initial arsenic Astot ≈ 1000 µg/L.
1993), though final arsenic removal was still greater than for pre-formed ferrihydrite (see
Section 2.4.5 for more discussion). In ECAR, arsenic concentrations at tM > 15 min
either stabilize at a constant value (suggested by kinetics test 1 in Figure 3.4), continue
to slowly decrease (at a rate that is below the Quick Test detection limit), or slowly
decrease and then begin to increase (as suggested by coprecipitation experiments). If
arsenic continues to decrease slowly, then very little is gained past tM = 15 minutes for
a large cost in treatment time (hours to days). If arsenic stabilizes or begins to increase,
then nothing is gained and some removal capacity may be lost. Thus tM was chosen to
encompass the period of rapid arsenic removal but exclude slow removal past a reasonable
treatment time of several hours. Since tM is only one portion of the treatment time, it
111
was set at 60 minutes for Matrix 1 batch tests1 to ensure a total treatment time of less
than 2 hours.
The mixing kinetics profile in ECAR is likely to change with current processing
time, tCP . Complexation with arsenic will begin as soon as iron (hydr)oxides begin
to form during electrolysis, and more or less of the rapid complexation phase seen in
Figure 3.4 (tCP = 5 min) may occur in this time. The kinetics profile may also change
with current density if the adsorbent generated by ECAR is different. For these reasons,
a check was placed on the tM value of 60 minutes for batch tests with different values
of tCP and current density. A 120 mL aliquot was removed during mixing after tM =
40 min for each batch test. The aliquot was filtered and tested for arsenic using the
Quick Test. If this measurement disagreed with a Quick Test measurement taken at tM
= 60 min by more than 20% (the approximate precision of the Quick Test), mixing was
allowed to continue for an additional 20 minutes to ensure that the rapid complexation
phase was complete. For Matrix 1 , all Quick Test measurements taken at tM= 40 min
and tM = 60 min agreed to within 20%.
It should be noted that in kinetics tests, the mixing time was defined as the time
between dosage completion (i.e. electrodes removed from water) and pouring out the
sample aliquot for filtration. Some additional complexation may have occurred during
the 3-4 minutes between the start and end of filtration, leading to an uncertainty in
complexation time of about 3 min. This is expected to be more significant during the
period of rapid complexation at tM < 10-15 min. It should also be noted that the1The only exception was the test at j = 1.1 mA/cm2 for tCP = 50 min, which has tM = 30 min. In
this case, the Quick Test measured Astot=0 µg/L at both tM = 20 min and tM = 30 min, indicating thatrapid complexation was complete.
112
electrochemical cell and SBGW-1 preparation procedure used for kinetics tests were
slightly different than those described above. The iron electrode had the same active
area, but was in a 3D coil shape rather than a spiral. The SBGW-1 water used was
spiked to Astot = 1000 µg/L instead of 600 µg/L (still containing a 50:50 ratio of As[III]
to As[V]). The pH of the water batch was lowered with small amounts of dilute sulfuric
acid in place of bubbling CO2 (pre-treatment pH was 6.85 and 6.93 for kinetics tests 1
and 2 respectively). None of these differences are expected to have any significant effect
on the kinetics behavior during mixing.
Replication
All tests were duplicated, and in some cases repeated 3-5 times. However,
later ICPMS results of the pre-treatment sample water showed a greater than expected
range of initial As[III]/Astot ratios for SBGW-1, in some cases falling much below 0.4
or greater than 0.6 (see Section 3.2.3). An extreme initial As[III]/Astot ratio is likely to
bias batch test results because, in general, As[III] requires a higher dose to remove than
As[V]. Indeed, all replicates with extreme initial As[III]/Astot ratios followed this trend
relative to the original tests. In order to maintain a consistent sample, batch tests with
a measured As[III]/Astot <0.40 or > 0.6 were removed. To account for water batches
with no measured As[III]/Astot, and given the general trend for As[III]/Astot to decrease
with storage time (measured in Figure 3.1), all tests using water batches > 1 day old
were removed as well. This cut left 7 out of 16 total batch tests with only a single
measurement (no repeat), 8 out of 16 with two measurements, and 1 test with three
113
measurements.
Variability estimates
To estimate the expected variability between outcomes of replicated batch tests
(i.e. batch tests performed using the same ECAR operating conditions to treat water
of the same composition), the eight Matrix 1 tests with at least one acceptable repeat
measurement were used. Outcomes between ”identical” batch tests may differ due to
one or more of the following: slight differences in composition between water batches,
slight differences in ambient pressure and temperature in the lab (expected to be small),
slight differences in initial DO content (measured standard deviation of 0.5 mg/L be-
tween water batches) or initial pH (standard deviation of 0.42 pH units between water
batches) as well as pH and DO drift during dosing and mixing (size of drift dependent
on current processing and mixing time), differences in the initial spiked arsenic concen-
tration (ICPMS measured standard deviation of total initial arsenic of 75 µg/L, or 13%)
and differences in the initial As[III]/As[V] ratio (ICPMS measured standard deviation
of 0.05, or 10%) and finally, uncertainty associated with ICPMS arsenic measurement
(quoted as within 10% by C&T lab). Many of the above sources of variability have
a complicated and likely non-linear effect on the final arsenic concentration measured
in a given batch test. Some are likely to cause more variability with increasing final
arsenic concentration, while some are expected to cause random variability. The avail-
able data set is too small to discern between a linear and non-linear model, and thus a
linear model has been chosen to obtain a rough estimate of variability. To remove error
114
associated with slightly different initial arsenic concentrations between repeated tests,
all arsenic concentrations were normalized to have an initial concentration of Asinit =
580 µg/L. Figure 3.5 shows the replicate variability between repeat measurements (i.e.
|y2 − y1|/√
2 for two replicate measurements y1 and y2)2 as a function of the average
final arsenic concentration.
0 50 100 150 200 250
020
4060
80
Average Final Arsenic Concentration ((µµ g L))
Rep
licat
e V
aria
bilit
y ((µµ
gL))
Figure 3.5: Replicate variability between arsenic concentrations measured in identicalbatch tests (repeat measurements) for Matrix 1 (where repeats are available) as a func-tion of the average final arsenic concentration, Asfave. The starred point has a higherthan expected standard deviation, and represents the batch test performed at j= 30mA/cm2 and tCP = 5 min. The dotted line in a linear fit to non-starred points.
Figure 3.5 indicates a trend (as expected) of increasing uncertainty with in-
creasing average arsenic concentration. The trend appears approximately linear with2For the one case with three acceptable replicate measurements, the replicate variability is the stan-
dard deviation.
115
the exception of the final point (designated by a star) corresponding to the batch test
performed at j= 30 mA/cm2 and tCP = 5 min. The deviation between tests for this point
is 84 µg/L, or 34% of the average final arsenic value, much higher than the deviation
measured for nearby points. While this deviation is consistent with highly non-linear
behavior, it is also consistent with an unknown procedural error in one or both of the
batch tests at j= 30 mA/cm2, tCP = 5 min. Attribution to a procedural error is further
supported by the fact that the resulting final arsenic concentrations for both batch tests
at j= 30 mA/cm2, tCP = 5 min were higher than the final arsenic concentrations at
j= 30 mA/cm2, tCP = 3 min, despite representing a higher dose. No other set of batch
tests violated the trend of a higher dosage leading to a lower final arsenic concentration3.
For these reasons, a procedural error has been assumed, and all batch tests for j= 30
mA/cm2, tCP = 5 min were discarded for further analysis.
The remaining points were fit to a linear model, shown in Figure 3.5 as a dotted
line. The resulting estimated variability for Matrix 1 points, σM1, as a function of final
arsenic concentration, Asf , is (in units of µg/L):
σM1 = 3.2± 4.1 + (0.032± 0.039)Asf (3.2)
The R2 value for this fit was quite low, with R2 = 0.1192. In addition, the estimated
errors for each parameter in Equation 3.2 are greater than 100%. However, no better
estimate is allowed by the small data set, leading to the adoption of this model over
none. The model fit is low compared to the reported ICPMS measurement errors of3In one other case (j=110 mA/cm2), the average of two tCP = 5 min batch tests resulted in a slightly
higher final arsenic concentration than the average of the tCP = 3 min, however the difference was wellwithin the error of ICPMS measurement and so not significant.
116
±5% with a reporting limit of 1 µg/L. This could be due to ICPMS error inflation by
C&T or the weakness of the fit.
Experiments with no external voltage applied
A series of tests were also run to study the arsenic removal capacity of the iron-
copper electrode system with no external voltage applied. These tests followed the same
procedure as above, but with the galvonostat removed from the circuit (i.e. no external
voltage applied). Two batch tests were performed with current processing times tCP =
11 and 50 min. Each test was repeated and the results averaged. The iron electrode
was rinsed in dilute HCl and DI water to remove rust between each batch test, with the
exception of the duplicate tCP = 50 min test. This test used a pre-rusted iron electrode
(rusted by allowing DI water to drip on the iron in air for several days). The measured
As[III]/Astot ratio of the water used was 0.62, slightly higher that the ratio allowed for
other Matrix 1 experiments. The measured current through the electrodes during these
tests ranged from 0.67 - 0.036 mA, usually starting high and progressively decreasing,
though in one case the current started low and increased. Two final batch tests were
performed with no external voltage and no electrodes in the water (i.e. j = 0 and tCP =
0). In essence, the synthetic groundwater was simply stirred for tM = 60 min and filtered,
giving a measurement of the removal capacity of background ions found in Bangladesh
groundwater (excluding iron) and filtration alone.
117
3.2.7 ECAR batch experiments run 2 - Matrix 2
Purpose
A second round of batch experiments was performed using SBGW-2 synthetic
groundwater (containing average levels of phosphate and silicate) and the 3 L-single
chamber electrochemical cell. The purpose of this run, designated Matrix 2 , was to
determine the minimum dose required to reduce 500-600 µg/L arsenic in water with
expected levels of phosphate and silicate to below the Bangladesh and WHO limits (50
µg/L and 10 µg/L respectively), and to expand the range of current densities observed.
The single chamber electrochemical cell was used in place of the separated cathode
electrochemical cell (used in Matrix 1 ) after internal testing revealed little to no difference
in arsenic removal capacity for the two cells operating under comparable conditions.
Procedural differences from Matrix 1
For Matrix 2 , a single batch experiment was run for each current density and
aliquots were removed at designated times corresponding to specific charge densities.
Aliquots of 50 mL were removed using a glass luer tip syringe from the top of the cell.
Sampling times were calculated according to Equation 2.32 taking into account the new
treatment volume after sampling. Current densities 1.1, 5.0, 10, 30, and 100 mA/cm2
were tested with aliquots removed after dosages of 25, 50 75, 100, 125, 150, and in
some cases 175 C/L. Current density 0.020 mA/cm2 was sampled after dosages of 6,
12, 18, 25, 50, and 75 C/L due to the higher observed removal efficiency. Each 50 mL
aliquot was mixed for tM > 60 min before filtration using a new membrane. If As[III]
118
Table 3.2: Operating parameters for Matrix 2 tests.
Current Operating Active Max MaxDensity Current Electrode Area Dosage tCP
a A/Vb
(mA/cm2) (mA) (cm2) (C/L) (min) (cm−1)
100 900 9 175 9.20 0.003030 900 30 150 8.13 0.01010 900 90 175 9.20 0.0305.0 500 100 175 16.33 0.0331.1 110 100 175 74.98 0.0330.020 3 150 75 1199.29 0.050
a tCP is the total current processing time for the batch test. Note that tCP for lower values of charge densitywill be lower. Note also that the treatment volume decreases slightly as aliquots are removed, decreasing therequired current processing time from that expected by Equation 2.31.
b Unless indicated otherwise, replicated tests used the same A/V ratio.
measurements were taken, the filtrate was immediately filtered through a speciation
cartridge (Section 3.2.1) and placed in a sample vial. As[III] was measured at least once
for each current density and dosage. Initial arsenic concentration was measured from
unfiltered aliquots removed before electrodes were placed in the cell. pH and DO were
measured 2-3 minutes before electrodes were inserted as well as immediately after dosing
during the mixing stage. Synthetic groundwater was made less than 24 hours prior to
testing.
Different sized electrodes bent to the same shape, and made of the same iron
wire, were used for these experiments in order to work within the current range of the
galvonostat. Operating current, current density, and active electrode area used for each
test are given in Table 3.2.
119
Mixing time kinetics profile
The mixing time for Matrix 2 tests was set at tM = 60 min based on the kinetics
tests from Matrix 1 (shown in Figure 3.4). To determine if the change in water batch
recipe or procedure affected the mixing kinetics profile, one kinetics profile was measured
in SBGW-2 using the Matrix 2 procedure. ECAR parameters were j = 1.1 mA/cm2,
q = 150 C/L, tCP = 66 min. The resulting profile is shown in Figure 3.6. Although
the arsenic concentration is still decreasing at tM = 60 min, the gain in percent arsenic
removal is only about 0.2% between tM = 60 - 100 min, or 0.005% removal per min. This
indicates that slow complexation has been reached by tM = 60 min even in SBGW-2.
20 40 60 80 100
68
1012
14
Mixing Time (min)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
9998
.598
97.5
% A
rsen
ic R
emov
ed
WHO limit
Figure 3.6: Mixing time kinetics profile for SBGW-2 spiked to Astot = 510 µg/L treatedat j = 1.1 mA/cm2, q = 150 C/L, tCP = 66 min. Residual arsenic concentration (leftaxis) and percent arsenic removed (right axis) as a function of mixing time are shown.A dashed horizontal line marks the WHO limit (10 µg/L).
120
Replication
All batch experiments, with the exception of j = 30 mA/cm2 and j = 100
mA/cm2, were repeated and the initial and residual arsenic concentrations at each value
of j and q were averaged. In addition, for the replicated batch tests, the maximum q
was increased from 150 C/L to 175 C/L, and hence there is only one observation per
current density at q = 175 C/L. With the exception of j = 0.02 mA/cm2, the repeated
tests used the same A/V ratio, and tCP associated with each charge density varied by
< 1 minute4.
Replicate Variability estimates
Variations in the arsenic concentration after ECAR treatment come from the
same sources in Matrix 2 experiments as in Matrix 1 (discussed in Section 3.2.6). How-
ever, procedural differences caused a decrease in the spread of measured initial arsenic
concentrations compared to Matrix 1 (for Matrix 2 , the average over all batch tests
was Asinit = 560 ±30 µg/L). The expected standard deviation as a function of arsenic
concentration (assuming a gaussian distribution) was estimated by fitting the replicate
variability (i.e. σ = |y2 − y1|/√
2 for two replicates) to the average final arsenic concen-
tration for those tests (Figure 3.7). The resulting estimated uncertainty for Matrix 2
points, σM2, as a function of final arsenic concentration, Asf , is (in units of µg/L):
σM2 = 0.6± 1.6 + (0.071± 0.007)Asf (3.3)
4Small variations in tCP between batch tests were due to varying the aliquot volumes between tests- see Equation 2.32 for the relationship between treatment volume (which is reduced as each aliquot isremoved) and tCP .
121
The R2 value for this fit was R2 = 0.875. The adjusted R2 for this fit was R2adjM2 =
0.867 (compared to an adjusted R2 of R2adjM1 = -0.057 for the Matrix 1 fit).
0 100 200 300 400 500
010
2030
4050
Average Final Arsenic Concentration ((µµ g L))
Rep
licat
e V
aria
bilit
y ((µµ
gL))
Figure 3.7: Deviation in the final arsenic concentration measured between replicatedbatch tests for Matrix 2 as a function of the average final arsenic concentration, Asfave.The deviation plotted between two measurement y1 and y2 is just σ = |y2−y1|/
√2. The
dashed line in a linear fit weighted by 1/As2fave.
The current processing time required to provide a given dosage is a function of
the operating current I and not the current density j (Equation 2.31). In the Matrix
2 experiments, the current was allowed to vary between current densities, in some case
leading to drastically different current processing times. To isolate the effect of current
density from the effect of current processing time on arsenic removal capacity, a single
batch test was performed at 1.1 mA/cm2 using the same current, I = 3 mA, and hence
the same current processing time, as the earlier batch tests performed at 0.02 mA/cm2.
Aliquots were removed in the same manner as before at times corresponding to q = 6,
122
12, 18, 25, 50, and 75 C/L (to match the samples taken during the 0.020 mA/cm2 batch
test). To maintain the current density, an electrode with 2.73 cm2 active area was used
(A/V = 0.00091 cm−1).
3.2.8 Adsorption using post-synthesis ECAR-generated adsorbent
Two batch experiments, PSA 1 and PSA 2, were performed to measure the
arsenic removal capacity of ECAR-generated iron (hydr)oxides particles allowed to form
and age before being exposed to arsenic. To obtain post-synthesis ECAR-generated (PS-
EGA) adsorbent, synthetic groundwater was prepared according to the SBGW-2 recipe
(with no initial arsenic) and dosed using the 3 L-single chamber electrochemical cell at j
= 5.0 mA/cm2, q = 175 C/L, and tCP = 17.3 min. After dosing, the solution was divided
into Sample A (2 liters) and Sample B (1 liter). In sample A, arsenic was added at the
start of mixing, and in Sample B, arsenic was added after 60 minutes of aging (while
stirred). In both cases, the solution was mixed an additional time tM after arsenic was
added, followed by filtration (note that tM does not include the 60 minutes of stirring
prior to arsenic addition for Sample B).
Arsenate and arsenite stock solutions were added sequentially to sample A and
sample B solutions (while stirring) in an amount corresponding to a final solution con-
centration of 300 µg/L As[V] plus 300 µg/LAs[III] (ratios of stock As[V] and As[III]
solution added to existing solution were 6 · 10−4 and 6 · 10−3 respectively). Stock solu-
tions were the same as those used to make water batches (see Section 3.2.3). During the
first batch experiment, unfiltered aliquots were removed to measure initial arsenic within
123
30 seconds of adding arsenic (HNO3 preservative was added right away to prevent pre-
cipitation and settling in the sample vial). However, the measured concentrations were
quite low compared to expected values (Astot = 210 and 300 µg/L for samples A and
B respectively), suggesting that the arsenic had not fully mixed within the solution at
the time of sampling. During the second batch experiment, aliquots to measure initial
arsenic were removed 60 minutes after arsenic was added to solution (again, the aliquots
were not filtered and preservative was added right away to prevent settling in the sample
vial). These results were much closer to the expected amount of arsenic (Astot = 630
and 590 µg/L for A and B respectively). Therefore, the initial arsenic level for both
samples A and B in Exp 1 were estimated to be the average of measured initial arsenic
for samples A and B in Exp 2 (Astot = 610 µg/L).
In Exp 1, aliquots were removed from the top (using a glass luer tip syringe)
of both samples A and B after tM = 20, 40, 60, 80, 100, 120 min. Aliquots were 40 mL
for all but tM = 60 min, when an aliquot of ≈ 500 mL was removed for XAS sample
preparation (Section 4.2.4). In Exp 2, only one aliquot of ≈ 500 mL was removed after
tM = 60 min. In both Exp 1 and 2, a mixing time tM=60 min was used to compare
to Matrix2 coprecipitation tests. After mixing for time tM = 120 min and tM = 60
min for Exp 1 and 2 respectively, unfiltered samples of each solution (A and B) were
placed in beakers and allowed to settle overnight. The samples were filtered and tested
after settling for 1 and 2 days (1 day only for Exp 2). This was done to determine if
further complexation and coagulation was still occurring. These points cannot properly
be taken as an extension tM since the solution was not being mixed.
124
3.2.9 Polarization studies
Voltammetry is a sensitive technique used to probe the electrochemical behavior
of a system. It involves adjusting the potential of an electrochemical cell and observing
the resulting current response or polarization curve (i.e an i-E curve, discussed in Section
2.5.15). The magnitude of the current flowing through the cell will be determined by
the rates of heterogeneous and homogeneous reactions occurring in each half-cell as well
as the rate at which reactants in solution can be transported to and from the electrodes
by diffusion, migration, or convection. If the potential is scanned at a predetermined
rate, analysis of the current response as a function of time can lead to useful information
about the different reactions occurring in a cell as well as the relative reaction rates.
If a three-electrode cell (discussed in Section 2.5.1) is used, then the working electrode
can be isolated and the current response will reflect reaction rates and mass transport
occurring at the working electrode-solution interface only.
Voltammetry is also extremely useful to determine if passivation is occurring on
an electrode. Passivation can occur when an oxide layer forms on the electrode surface,
preventing further electron transfer. This can appear very prominently on an i-E curve,
as the net current will decrease to zero when passivation is complete.
Polarization curves were measured for the iron wire electrode used in Matrix 1
and Matrix 2 experiments using a 60 mL three-electrode cell. The three-electrode cell
was fabricated as depicted in Figure 3.8 including a platinum counter electrode (surface
area 1.07 cm2) and an Ag/AgCl (3M NaCl) single junction reference electrode (5.7 mm5This text relies heavily on background presented in Section 2.5.1, and it will be assumed here that
the reader is familiar with its contents
125
outer diameter) with a vycor frit (Bioanalytical). The working electrode was fashioned
from iron wire by covering the length of the wire in epoxy resin and filing one end to
expose a disk of iron with a diameter of 0.18 cm (surface area 0.025 cm2). The exposed
end was bent to be within 1 mm of the reference electrode tip to reduce the IR potential
drop (uncompensated resistance) due to the electrolyte (see Figure 3.8). Solution was
purged with N2 gas for approximately 2 min before conditioning for each scan and N2 gas
was used to purge the headspace during conditioning and scanning. The dissolved oxygen
levels at the end of any given scan were < 0.10 mg/L. The entire apparatus was placed
inside a grounded copper mesh faraday cage to reduce electrical interference (primarily
needed for EIS experiments, as explained in Section 3.2.10). Scans were performed with
a Gamry Instruments Femptostat CE scanning potentiostat running Gamry Framework
software v4.35.
The Ag/AgCl reference electrode was tested against a second saturated calomel
reference electrode (Beckman; ceramic frit) before use. The vycor frit on the reference
electrode was secured to the glass electrode body using heat shrink wrap that was rela-
tively easy to loosen, potentially allowing ions to bypass the vycor frit. To see if this was
the case, the electrical impedance of the reference electrode was measured with only the
vycor frit submerged and compared to the impedance with the potentially lose shrink
wrap seal submerged. The impedance did not change, and so it was assumed that ions
were not able to bipass the vycor frit through the shrink wrap seal.
The working electrode was sanded using Silcon Carbide 1500 mirror fine wa-
terproof sandpaper before each scan. It was later suggested that non-reproducible back-
126
N Inlet2
Air Outlet
PlatinumCounterElectrode
Ag/AgClReferenceElectrode
Exposed IronWorkingElectrode
EpoxyCoating
Stir Bar
CopperFaradayCage
Counter Electrode Lead
Reference Electrode Lead
Working Electrode Lead
Ground
ScanningPotentiostat
Figure 3.8: Schematic of Three-Electrode Cell used for polarization studies and Elec-tronic Impedance Spectroscopy (EIS). The cell contained a platinum counter electrode,an Ag/AgCl reference electrode, and an exposed iron disk working electrode (surfacearea 0.025 cm2). The cell is described more fully in the text (Section 3.2.9).
127
ground currents could be reduced if the electrode surface was systematically polished
before each test using alumina or diamond paste as the polishing material. Background
currents did not appear to be an issue for this round of tests, but more rigorous polishing
should be adhered to in the future.
Polarization curves for the iron wire electrode were measured at a scan rate of
ν = 0.1 mV/s in 0.1M KClO4 (Alfa Aesar) added to DI water and 0.1M KClO4 added
to synthetic Bangladesh groundwater (SBGW-2 recipe, see Table 3.1). The perchlorate
salt was added to reduce the IR-drop between the reference and working electrode6.
Perchlorate was chosen because it dissociates substantially at neutral pH, the ClO –4 ions
are slow to oxidize at room temperature in the voltage range scanned, and the ions do
not form complexes with metal. Before each scan, the iron electrode was conditioned
at a constant current of - 1nA for 6 hours. Open circuit voltage for plain perchlorate
and perchlorate plus SBGW-2 was -0.570V and -0.630V vs Ag/AgCl respectively. Scans
started at the open circuit voltage scanning up to -0.1V, down to -1.6 V and back up to
-0.1V. Oxidation curves were derived from the second pass between open circuit voltage
and -0.1 V. This ensured that any passive layer which might have formed on the iron
electrode while exposed to air was completely reduced before the polarization curve.
The pH of the solution were measured before the initial N2 purge as well as
after the completion of the scan. For plain perchlorate solution, the initial pH was pHi
= 6.80 and the final value, pHf = 9.60. For perchlorate plus SBGW, the values were
6DC transient voltammetry requires concentrations above 0.1 mM in order to achieve a significantFaradic current signal relative to background. Trace concentrations down to 10−7 M may be determinedusing AC, pulse, or square-wave methods to enhance the Faradic-to-background current ratio (Bond,2002).
128
pHi = 7.62 and pHf = 9.14. The majority of this change likely occurred during the
initial N2 purge and before the start of the scan, however, there is no way to be sure7.
3.2.10 Electrochemical Impedance Spectroscopy
Interpreting the results of Electrochemical Impedance Spectroscopy scans is
based on the assumption that an equivalent circuit can be defined for an electrochemical
system and the basic form of the equivalent circuit can be determined from the system’s
impedance response when driven at different frequencies. The impedance, Z, is measured
by exciting the system with a potential E(t) = E0cos(ωt) and measuring the response
current I(t) = I0cos(ωt − φ). The current magnitude and phase shift measured in the
response give the complex impedance at frequency ω according to:
Z(ω) =E
I= Z0(cos(φ) + isin(φ)). (3.4)
If Z is measured for a series of different excitation frequencies, the real and imaginary
parts can be plotted on a Nyquist plot such as that shown in Figure 3.9. Note that
the Y-axis is negative and that each point on the Nyquist plot is the impedance at one
frequency, ω. The Nyquist plot shown is for a simple equivalent circuit with a single
time constant, showing a characteristic semi-circle. EIS plots are often made up of
numerous time constants, and often only a portion of the semi-circle created by a single
time constant is measured. For a given electrochemical system, the measured Nyquist
plot is compared or fit to the expected Nyquist plot that would result from an equivalent7PH could not be measured immediately after the initial N2 purge because this would allow oxygen
back into the chamber.
129
circuit. Knowledge of the reaction pathways in the real system in inferred from matching
the behavior of different pathways to the behavior of equivalent circuit elements.
-ImZ
RealZ0
|Z|
φω = 0ω = 8
ω
Figure 3.9: A simple Nyquist plot showing complex impedance vector Z(ω). Note thatthe Y-axis is negative and that each point on the Nyquist plot is the impedance at onefrequency, ω. In this plot, low frequencies appear on the right side and high frequenciesappear on the left. This is the case when the impedance falls as the frequency rises(usually, but not always, true).
Equivalent circuit elements must have a basis in electrochemical theory to be
useful. Some of the common elements found in a three-electrode electrochemical cell
include basic resistance and capacitance. Resistances include electrolyte resistance be-
tween the working and reference electrode, polarization resistance as the potential is
changed which includes the kinetics of reactions and rates of diffusion to and from the
electrode, and charge-transfer resistance based on single kinetically controlled reactions
in equilibrium. Capacitance is caused by the electrical charge that gathers at the elec-
trode surface to compensate for charge in the surrounding electrolyte (forming what
is called the double layer capacitance). Diffusion can lead to impedance as well since,
at high frequencies, the ions don’t have to move very far to respond, whereas at low
130
frequencies, they must diffuse farther. Impedance due to diffusion is known as War-
burg impedance. Other more complicated electrochemical cell behavior can be modeled
using more complicated circuit elements as well, making EIS a powerful technique for
understanding cell behavior.
In practice, this technique suffers from some important limitations. For exam-
ple, impedance analysis is much easier in linear systems (circuits) than non-linear ones.
However, electrochemical systems are highly non-linear (doubling the voltage in general
will not double the current). Thus in normal EIS practice, only a small (1 to 10 mV)
AC signal is applied to confine the measurement to a pseudo-linear segment of the cell’s
current versus voltage curve. A second limitation is that the system must be in steady
state over many hours to take an EIS spectrum. In practice, this is very difficult to
achieve and may lead to very inaccurate results. Finally, this technique is also highly
dependent on the equivalent circuit model chosen by the researcher. Numerous models
may fit the same impedance response, and additional knowledge of the system is required
to chose the most appropriate one.
EIS scans were performed to measure the electrolyte resistance in synthetic
Bangladesh groundwater (SBGW-1). The same three-electrode cell shown in Figure 3.8
was used (described in the text of Section 3.2.9) hooked up to a Gamry Instruments
Femptostat CE scanning potentiostat running Gamry Framework software EIS300 v.4.35.
Before each scan, the current was monitored until approximate steady state was reached
(usually 40 - 120 seconds). AC potential with an amplitude of 10 mV was applied around
the open circuit voltage (-0.65 V) in a frequency range of 1-100,000 Hz (10 data points
131
per decade). It was found through trial and error that the results were much less noisy
when the three-electrode cell was surrounded by a grounded copper mesh Faraday cage,
so the cage was used for all measurements. The solution was 68 mL of freshly prepared
synthetic Bangladesh groundwater (SBGW-1 recipe) with initial pH = 7.00 and DO =
1.53 mg/L (solution was purged with N2).
The setup and theory was tested using a dummy cell, which is just a real
circuit containing the components shown in Figure 3.11. The dummy circuit values
were Re = 200Ω and Rct = 3000Ω. The EIS scan showed a nice semi-circle shape with
low frequency intercept at Zreal(ωlow) = 3210 ± 20Ω and a high frequency intercept at
Zreal(ωhigh) = 201± 3Ω, correctly recovering the values of Rct +Re and Re respectively,
indicating that the setup was sound and correctly calibrated to within 2%.
3.2.11 Sedimentation tests
Sedimentation was compared to 0.1µm vacuum filtration as a means to separate
ECAR generated particles from potable water during both the Matrix 1 and Matrix 2
batch experiments. During Matrix 1 , an additional batch test was performed for each
of four experiments with different operating conditions (including the following: (1) j
= 1.1 mA/cm2, q = 23 C/L, tCP = 3 min; (2) j = 1.1 mA/cm2, q = 85 C/L, tCP = 5
min; (3) j = 0.07 mA/cm2, q = 25 C/L, tCP = 50 min; and (4) j = 0.70 mA/cm2, q
= 15 C/L, tCP = 3 min). Dosed water from each batch test (850 mL) was placed in a
1-liter beaker and allowed to sit quiescently8 (no additional mixing was performed). At8For this work, quiescent refers to sitting without mixing or shaking of any kind.
132
designated times, small aliquots were removed from the top of the beaker using a luer
tip syringe and tested for arsenic.
A brief test was conducted to determine if magnets could accelerate the sepa-
ration of ECAR- generated particles from potable water. The mixed effluent from two
Matrix 1 tests was placed in two small beakers - one with powerful flat magnets (sal-
vaged from computer hard drives) taped to the bottom and one with no magnets. The
two beakers were allowed to sit and visually evaluated one day later. Aliquots from the
top of the beaker were testing using the QuickTest, but these values were not verified by
ICPMS.
During Matrix 2 experiments, SBGW-2 water (Matrix 2 procedure) was used
that had been dosed at j = 1.1 mA/cm2, q=100 C/L, and tCP = 44.7 min. One portion
of the dosed water was mixed for 30 minutes using a magnetic stirrer and set aside, while
the rest was mixed for 60 minutes, providing two samples for experimentation to see if
prior mixing could accelerate the settling process. After mixing, samples sat quiescent
and uncovered in 1-liter glass beakers to allow settling. In each sample, aliquots were
removed from the top 2 cm using glass luer tip syringe at intervals from 0 to 45 hours
after mixing ceased. Half of the aliquot was filtered using a 0.1µm membrane before
arsenic testing and half the sample was not filtered.
133
3.3 Results and Discussion
3.3.1 Resistivity of synthetic Bangladesh water
Electrochemical Impedance Spectroscopy (EIS) was used to measure the resis-
tivity of synthetic Bangladesh groundwater (SBGW-1 recipe). Several EIS scans were
performed about the open circuit potential in freshly prepared SBGW-1 water (full pro-
cedure described in Section 3.2.10). The high frequency portion (near 100 kHz) of a
representative scan (Figure 3.10) reveals half of the semi-circle shape characteristic of
a single time constant (similar to the example shown in Figure 3.9). A simple equiv-
alent circuit with a single time constant was used to model the system. This circuit,
shown schematically in Figure 3.11, includes a double-layer capacitance, Cdl, in parallel
with resistance due to charge-transfer and polarization, Rct, which occurs across the
double layer (the electrochemical meaning of these elements is explained more fully in
Section 3.2.10). Cdl and Rct are added in series to the resistance of the electrolyte, Re
between the working and reference electrode.
An EIS scan of the circuit shown in Figure 3.11 would produce a semi-circle
on a Nyquist plot crossing the real axis at two points. At very high frequencies, the
capacitor will contribute very little to the impedance Z, and hence most of the current
will flow through the capacitor Cdl and resistor Re, bypassing the resistor Rct. Thus
the high frequency real axis intercept will correspond to the resistance9 Zreal(ωhigh) =
Re. Using the circuit in Figure 3.11 as a model, the bulk resistance of the electrolyte9Conversely, at low frequencies, the impedance of the capacitor is very high and most current will
flow through the resistors Rct and Re in series. The real value of Z at this point should be Zreal(ωlow) =Rct +Re, however, this does not concern the current measurement.
134
1400 1600 1800 2000 2200
020
040
060
080
0
Real Impedance ((ΩΩ))
−−Im
agin
ary
Impe
danc
e ((ΩΩ
))
Figure 3.10: A Nyquist plot for a representative electrochemical impedance scan aboutthe open circuit potential in synthetic Bangladesh groundwater (SBGW-1 recipe).
Cdl
Rct
Re
Figure 3.11: Equivalent circuit used in EIS analysis to measure electrolyte resistance ofsynthetic Bangladesh groundwater. Components include the double-layer capacitance,Cdl, the charge-transfer resistance, Rct, and the electrolyte resistance, Re.
135
should correspond to the high frequency impedance real axis intercept measured from
Figure 3.10.
The high frequency impedance real axis intercept averaged over 4 separate EIS
scans is 1420±180 Ω. This corresponds to an electrolyte resistance of Re = 1420±180Ω.
The resistance is measured across a gap of 1 mm over an electrode area of 0.025 cm2,
leading to a resistivity of ρe = 355Ω·cm. As expected from the composition of SBGW-1,
the resistivity is very high. However, it is still not as high as the 1000-2000 Ω · cm
reported for groundwater in Bangladesh (Burgess & Ahmed (2006), converted from a
conductivity of 500 - 1000 µS/cm). This could partially be due to the excess sulfate
in SBGW-1 compared to real Bangladesh groundwater (sulfate was reduced from 81 to
8 mg/L in SBGW-2 to be more consistent with real values). In either case, the resistivity
is high enough to suggest that the voltage required of constant current treatment may
be high in Bangladesh if no modifications are made. A small amount of table salt (below
the taste threshold) added to the water before treatment could potentially increase the
conductivity and lower the power consumption of treatment - this should be explored in
future research.
It is important to note that only a small fraction of the power of EIS has been
employed in this measurement. EIS could be very useful to determine the reactions occur-
ring during ECAR treatment. For example, EIS scans around different mean potentials
would likely be able to distinguish whether the iron electrode in ECAR is producing
Fe 2+ leading to Fe[II]hydr(oxides) that oxidize to Fe[III]hydr(oxides), or Fe 3+ leading to
Fe[III]hydr(oxides) directly (the former would be associated with additional resistance).
136
This was the original intention for applying it to this work, however time constraints did
not allow EIS to be fully utilized. This single measurement and detailed methods section
(Section 3.2.10) can act as a guide for the student or students who carry on and expand
on this work. This measurement does verify a working setup and capability to continue
EIS scans in the future. Suggested future measurements include (1) EIS scans around
a series of potentials associated with the current densities used in Matrix 1 and Matrix
2 to determine if a different chemical pathway (characterized by a different impedance)
predominates at different current densities, and (2) a bulk resistance measurement of
the SBGW-2 synthetic and real groundwater in ECAR prototype units.
3.3.2 Arsenic removal capability in synthetic Bangladesh water
It is important that any technology suitable for Bangladesh be able to reduce
relevant arsenic levels below either the legal maximum in Bangladesh (50 µg/L), or
the more conservative maximum recommended limit by WHO (10 µg/L) in Bangladesh
groundwater. Batch tests of ECAR treatment have been performed over a range of charge
and current densities (q and j) using two recipes for synthetic Bangladesh groundwater
- SBGW-1 and SBGW-2 (composition described in Table 3.1). SBGW-1 and SBGW-2
differed in the amount of phosphate and silicate added; SBGW-1 had very low levels of
phosphate and silicate while SBGW-2 had average levels of both ions relative to tube
wells in Bangladesh. In each case, SBGW was spiked with 300 µg/L As[III] and 300 µg/L
As[V] (Astot = 600 µg/L).
Residual arsenic concentrations after batch tests in SBGW-1 are listed in Ta-
137
Table 3.3: Results of ECAR batch tests in synthetic Bangladesh groundwater (SBGW-1) containing low phosphate and no silicate. Initial and residual arsenic concentrationsare averaged over replicate batch tests. Boldface indicates the lowest residual arsenicconcentration achieved for each current density. Errors on arsenic measurements are10%.
Current Charge Current Initial Residual % As NumDensity Density Processing Time As As Removed Replicates
(mA/cm2) (C/L) (min) (µg/L) (µg/L)
0a 0a 0 580 530 8 20b 0b 11c 580 200 65 20b 0b 50c 580 93 84 2
0.07 1.5 3 560 370 34 10.07 2.5 5 590 220 64 20.07 5.4 11 600 160 73 10.07 25 50 620 19 97 2
0.30 6.4 3 580 170 71 10.30 11 5 590 260 57 20.30 23 11 580 130 78 10.30 110 50 560 6.4 99 2
0.70 15 3 600 160 74 20.70 25 5 590 84 86 20.70 54 11 530 29 95 10.70 250 50 580 5.7 99 1
1.1 23 3 580 60 90 31.1 39 5 530 54 90 11.1 85 11 620 8.5 99 21.1 390 50 580 4.3 99 1
a In this case, electrodes were never in contact with the solution, though the solution was stirred for tM =60 min and filtered, similar to other batch tests.
b For these tests no external voltage was applied, but the electrodes were in contact and some small current(0.0004 - 0.007 mA/cm2) flowed through the electrodes due to natural electrochemical processes.
c In this case, current processing time represents the amount of time that the electrodes were in contact withsolution, though no external voltage was applied.
138
0 100 200 300 400
010
020
030
040
050
060
0
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
8060
4020
0
% A
rsen
ic R
emov
ed
1.1 mA cm2
0.7 mA cm2
0.3 mA cm2
0.07 mA cm2
Bangladesh Limit
WHO Limit
0 50 100 150 200 250
050
100
150
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
9590
8580
7570
% A
rsen
ic R
emov
ed
1.1 mA cm2
0.7 mA cm2
0.3 mA cm2
0.07 mA cm2
Bangladesh Limit
WHO Limit
Figure 3.12: Arsenic removal (concentration on left vertical axis; percent arsenic removedon right vertical axis) as a function of charge density, q, after ECAR treatment atcurrent densities 0.07 - 1.1 mA/cm2 of low phosphate, zero silicate synthetic Bangladeshgroundwater (SBGW-1 recipe - Matrix 1 conditions). Lower plot presents same dataas upper plot on a reduced vertical scale to show detail at low arsenic concentrations.Percent arsenic removed is based on the average initial arsenic concentration. The WHOand Bangladesh maximum allowable arsenic limits are shown as dashed and dotted linesrespectively. Error bars represent the variability between identical batch tests.
139
Table 3.4: Results of ECAR batch tests in synthetic Bangladesh groundwater SBGW-2, including the average concentrations of phosphate and silicate found in Bangladesh.Initial and residual arsenic concentrations are averaged over replicate batch tests. Bold-face indicates the lowest residual arsenic concentration acheived for each current density.Errors on arsenic measurements are 10%.
Current Charge Current Initial Residual % As NumDensity Density Processing Timea As As Removed Replicates
(mA/cm2) (C/L) (min) (µg/L) (µg/L)
0.02 6 97.5 580 260 56 10.02 12 195.5 580 58 90 10.02 18 291 580 16 97 10.02 25 407.27 560 9.7 98 20.02 50 805.54 560 4.9 99 20.02 75 1199.29 560 6 99 20.02 100 1638.88 540 4.2 99 10.02 125 2041.67 540 2.8 99.5 10.02 150 2441.67 540 1.8 99.7 11.1 25 11.24 550 130 77 21.1 50 22.37 550 57 90 21.1 75 33.38 550 33 94 21.1 100 44.27 550 20 96 21.1 125 55.03 550 14 97 21.1 150 65.67 550 8.2 98 21.1 175 74.98 590 6.0 99 15 25 2.48 520 150 71 25 50 4.92 520 75 86 25 75 7.33 520 46 91 25 100 9.7 520 33 94 25 125 12.07 520 23 96 25 150 14.39 520 14 97 25 175 16.33 540 10 98 110 25 1.38 560 160 72 210 50 2.72 580 72 88 110 75 4.08 560 55 90 210 100 5.42 560 36 93 210 125 6.73 560 27 95 210 150 8.04 560 18 97 210 175 9.2 580 11 98 130 25 1.38 530 160 71 130 50 2.75 530 80 85 130 75 4.12 530 47 91 130 100 5.47 530 33 94 130 125 6.8 530 24 95 130 150 8.13 530 14 97 1100 25 1.37 570 170 71 1100 50 2.72 570 84 85 1100 75 4.05 570 57 90 1100 100 5.37 570 40 93 1100 125 6.67 570 30 95 1100 150 7.95 570 12 98 1100 175 9.2 570 13 98 1
a As explained in Section 3.2.7, the current processing time differs slightly from that expected by Equation 2.34or 2.31 due to small solution volume changes as aliquots were removed for testing (changing A/V slightly).The current processing time associated with each charge density was instead calculated using Equation 2.32and the solution volume as a function of time (based on the aliquote removal schedule). Replicate experimentssometimes followed slightly different aliquot removal schedules, leading to slightly different (< 4%) currentprocessing times between tests. The average current processing time is listed for each test.
140
0 25 50 75 100 125 150 175
010
020
030
040
050
060
0
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
8060
4020
0
% A
rsen
ic R
emov
ed
0.02 mA cm2
1.1 mA cm2
5 mA cm2
10 mA cm2
30 mA cm2
100 mA cm2
Bangladesh Limit
WHO Limit
0 25 50 75 100 125 150 175
020
4060
8010
0
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
9590
85
% A
rsen
ic R
emov
ed
0.02 mA cm2
1.1 mA cm2
5 mA cm2
10 mA cm2
30 mA cm2
100 mA cm2
Bangladesh Limit
WHO Limit
Figure 3.13: Arsenic removal (concentration on left vertical axis; percent arsenic removedon right vertical axis) as a function of charge density, q, after ECAR treatment at currentdensities 0.02 - 100 mA/cm2 of average phosphate, average silicate synthetic Bangladeshgroundwater (SBGW-2 recipe - Matrix 2 conditions). Lower plot presents same dataas upper plot on a reduced vertical scale to show detail at low arsenic concentrations.Percent arsenic removed is based on the average initial arsenic concentration. The WHOand Bangladesh maximum allowable arsenic limits are shown as dashed and dotted linesrespectively. Error bars represent the variability between identical batch tests.
141
ble 3.3. For j = 0.30 - 1.1 mA/cm2, the final concentration is < 7 µg/L (99% arsenic
removal), reaching < 5 µg/L in the case of j = 1.1 mA/cm2. This demonstrates the
ability of ECAR treatment to reduce arsenic below the WHO limit of 10 µg/L in syn-
thetic Bangladesh groundwater with low phosphate and silicate (SBGW-1). Treatment
at j = 0.07 mA/cm2 did not reach the WHO limit after passing a charge density of q =
25 C/L (Asfinal = 19 µg/L). However, looking at the behavior of arsenic concentration
as a function of charge density (Figure 3.12; j = 0.07 mA/cm2 data points connected
with a line for emphasis), it seems likely that the WHO limit would be reached with the
addition of more charge.
Results of ”blank” tests (i.e. j ≈ 0 mA/cm2) are also listed in Table 3.3. For
blank tests, electrodes were in contact with the water for a time tCP , but not external
voltage was applied. In the case of tCP = 0 min, there was no electrode contact and
treatment went directly to the mixing stage (tM = 60 min) followed by filtration. In this
case, only 8% of the arsenic was removed, reaching a final concentration well above the
Bangladesh legal limit of 50 µg/L (Asfinal = 530 µg/L). This demonstrates the inability
of filtration alone to remove arsenic oxyanions from groundwater. Electrode contact for
tCP = 50 min was also insufficient to reach either the Bangladesh or WHO limit (Asfinal
= 93 µg/L), though 84% of the arsenic was removed. This removal was likely due to
EGA generated from natural rates of corrosion. Current during treatment fluctuated
from 0.0004 - 0.007 mA/cm2, indicating a positive corrosion rate. The much higher
removal achieved after treatment at j = 1.1 mA/cm2 demonstrate the effectiveness of
ECAR over technologies based on natural rates of rusting.
142
Residual arsenic concentrations after batch tests in SBGW-2 are listed in Ta-
ble 3.4. Final arsenic concentrations for j = 0.02 - 5 mA/cm2 are ≤ 10 µg/L, reaching
< 2 µg/L for j = 0.02 mA/cm2. This demonstrates the ability of ECAR treatment to
reduce arsenic levels to < 2 µg/L - well below the WHO limit of 10 µg/L - in synthetic
Bangladesh groundwater with relevant levels of phosphate and silicate. This demonstra-
tion is more significant than SBGW-1 because the SBGW-2 is more representative of
Bangladesh groundwater, specifically with respect to solutes that compete with arsenic.
Given the high levels of competitive ions in SBGW-2 (including more phosphate and
silicate combined than 80% of contaminated wells in Bangladesh - see Section 3.2.3),
these results strongly suggest that ECAR will perform equally well in Bangladesh.
For j = 10 - 100 mA/cm2 treatment in SBGW-2 (Table 3.4) arsenic concen-
trations were quite close to the WHO limit at the highest charge density, q = 175 C/L
(Asfinal < 15 µg/L). Looking at the arsenic concentration as a function of charge density
for j = 10 - 100 mA/cm2 (Figure 3.13), one sees that concentration is still falling at the
end of each run. This suggests that the WHO limit would be reached with the addition
of some small amount of charge density.
The arsenic removal capability of ECAR in groundwater with bicarbonate, sil-
icate, and phosphate as well as both As[III] and As[V] is extraordinary. Most chemical
adsorbents used in filters in Bangladesh report reducing arsenic to 20 - 50 µg/L, below
the Bangladesh legal limit but above the WHO recommended safe limit. These results
demonstrate that ECAR can be an effective even in regions of Bangladesh with high
levels of phosphate and silicate as well as As[III].
143
3.3.3 Effect of charge density
Charge density controls the amount of iron dissolved into solution during the
dosing stage of ECAR, and hence the amount of EGA generated. Figures 3.12 and 3.13
show the residual arsenic concentration as a function of charge density from Matrix 1
and Matrix 2 batch experiments respectively. It is apparent from both figures that more
charge density leads to more arsenic removal. It is also apparent that the arsenic removal
capacity (i.e. the slope of Figure 3.12 or 3.13) decreases significantly with the residual
arsenic concentration. This behavior is likely due a decrease in adsorption capacity of iron
(hydr)oxides at low arsenic concentrations (recall that iron (hydr)oxides typically obey a
Langmuir isotherm, discussed in Section 2.4.1). Recall however that the operating costs
depend on the average arsenic removal capacity over treatment, not the instantaneous
removal capacity. Thus it is the total charge requirement to meet either the Bangladesh
or WHO limit that is relevant for Bangladesh. The total charge density required to reach
the Bangladesh and WHO limit (qBDesh and qWHO respectively) has been estimated in
Table 3.5 based on Figures 3.12 and 3.13. In cases where data did not extend far enough
in charge density for arsenic to reach the WHO limit, nearby points (i.e. < 9 µg/L
above the WHO limit) were linearly extrapolated from the last two or three measured
data points. In all cases, extrapolation was < 20% of the distance between the last two
data points. The value of qWHO was used to calculate the arsenic removal capacity and
iron capacity as Fe 3+ (i.e. µg arsenic removed per mg of iron dissolved assuming all
iron enters solution as Fe 3+) in Table 3.5. Also calculated is an estimate of the current
processing time required to reduce arsenic to the WHO limit based on Equation 2.34
144
Table 3.5: Performance metrics to reach the Bangladesh arsenic limit (50 µg/L) andthe WHO arsenic limit (10 µg/L) based on ECAR batch tests in synthetic Bangladeshgroundwater containing a 1:1 ratio of As[III]:As[V]. Arsenic removal capacity, iron ca-pacity (assuming all is dissolved as Fe 3+), and total treatment time have been calculatedbased on the WHO limit and data in Figures 3.12 and 3.13.
Based on WHO limitTotal
Water As As Removal Iron Capacitye TreatmentType j qBDesh
a qWHOb Removedc Capacityd as Fe 3+ tCP
f tMg Timeh
(mA/cm2) (C/L) (C/L) (µg/L) (µg-Asrem (µg-Asrem (min) (min) (Hrs)/C) /mg-Fe 3+)
Matrix 1SBGW-1 1.1 50 70 570 8.1 42 8.99 60 1.15SBGW-1 0.70 50 70 560 8.0 42 14.12 60 1.24SBGW-1 0.30 50 70 570 8.1 42 32.96 60 1.55SBGW-1 0.07 15 25i 580 22j 116j 52.46j 60 1.87
Matrix 2SBGW-2 100 90 180i 560 3.0j 16j 10.21j 60 1.17SBGW-2 30 75 160i 520 3.3j 17j 8.89j 60 1.15SBGW-2 10 75 180i 560 3.1j 16j 9.92j 60 1.17SBGW-2 5.0 75 175 510 2.9 15 17.68 60 1.29SBGW-2 1.1 55 150 550 3.7 19 68.87 60 2.15SBGW-2 0.02 15 25 560 22 116 416.67 60 7.94
Otherk
SBGW-2 1.1 25 50 560 11 58 832.50 60 14.88
a Charge density required to reduce arsenic to the Bangladesh legal limit of 50 µg/L.b Charge density required to reduce arsenic to the WHO limit of 10 µg/L.c Arsenic removed to reach the WHO limit (10 µg/L) based on the average initial arsenic concentration.d Arsenic removal capacity is the average µg arsenic removed per Coulomb to reduce arsenic to the WHO limit
(10 µg/L).e This is the average amount arsenic removed per Fe 3+ added to the water to reduce arsenic to the WHO limit
(10 µg/L), assuming that all charge is used to oxidize Fe(0) to Fe 3+ (based on faraday’s law - Equation 2.19).f tCP is the estimated current processing time required to reach qWHO based on Equation 2.34 and the same
experimental conditions (i.e. the same A/V).g The mixing time, tM , is the same as that used for batch tests since the value of qWHO is based on batch test
data (and hence based on the choice of tM ).h Total ECAR treatment time to reach the WHO limit (10 µg/L) assuming tfilt = 0 min. Total treatment
time is Ttreat = tCP + tM + tfilt.i This value was linearly extrapolated past existing data using the last two or three data points. In each case,
the extrapolated distance was < 20% of the distance between the last two points and the final measuredconcentration was within 10 µg/L of the WHO limit.
j Based on data extrapolated just beyond the final measured batch test (see previous note).k Slow tCP experiment, described in Section 3.3.4.
145
(plugging in qWHO and the same value of A/V used in the original batch test). Total
treatment time to reach the WHO limit is calculated using the same value of tM used in
batch tests (since the determination of qWHO from batch tests depends in part on this
parameter).
For Matrix 1 tests in SBGW-1, the charge density required for removal to the
WHO limit falls between qWHO = 25 - 70 C/L depending on current density. For j =
0.30 - 1.1 mA/cm2, the charge density requirement is the same, and indeed, removal
appears to follow a very similar path for these current densities in Figure 3.12. In this
current density range, charge density is a more important factor in removal than current
density, agreeing with the findings of Chen et al. (2000) and Kumar et al. (2004). Arsenic
removal capacity in this region is around 8 µg-Asrem/C. It is difficult to interpret the
much higher removal capacity of 22 µg-Asrem/C extrapolated from j = 0.07 mA/cm2
data. From the upper plot of Figure 3.12 (data for j = 0.07 macmsq connected with
lines for emphasis), it appears that the j = 0.07 mA/cm2 data would follow nicely along
the path of other current densities if not for the last data point near q = 25 C/L. From
Table 3.3, we see that only one batch tests at this point is available. Further batch tests
at higher charge density are needed to determine if current density is affecting removal
at j = 0.07 mA/cm2.
For Matrix 2 tests in SBGW-2, the charge density required to reach the WHO
limit varies from qWHO = 25 - 180 C/L. Arsenic removal capacity is similar for current
density range j = 1.1 - 100 mA/cm2. However, at the lowest current density, j =
0.02 mA/cm2, removal capacity is 5-6 times higher. Unlike the case of j = 0.07 mA/cm2
146
in SBGW-1, data for j = 0.02 mA/cm2 extends far above the charge density required
to reach the WHO limit (Figure 3.12). This clearly demonstrates that charge density is
not the only predictor of arsenic removal capacity in ECAR. In light of this behavior,
it is likely that the increase in removal capacity calculated from extrapolated values for
j = 0.07 mA/cm2 is real, and removal capacity is progressively larger at lower current
densities.
3.3.4 Effect of current density
As discussed in Sections 1.4.3 and 2.5.5, investigators have reported mixed
results for the influence that current density has on the efficiency of electrocoagulation
processes. For arsenic removal with iron electrodes, Kumar et al. (2004) and Gomes et al.
(2007) reported no change in removal efficiency over a current density range of j = 0.65 -
1.53 mA/cm2 (pH 6-8) and j = 3 - 30 mA/cm2 (pH 2.4) respectively. For the removal of
other contaminants, Pouet & Grasmick (1995) reported that current density can affect
treatment efficiency, while Chen et al. (2000) disagreed, attributing the discrepancy to
the fact that charge density was not held constant in experiments of Pouet and Grasmick.
All of the above researchers have noted that current density affects treatment time.
As stated in Section 3.3.3, the batch test results presented in Figures 3.12 and
3.13 and performance metrics based on this data (Table 3.5) indicate that current density
does have some effect on arsenic removal capacity, though no effect is seen within certain
current density ranges. Recall from Section 2.5.2 that the current density is expected
to control the composition of iron (hydr)oxides in EGA. Because the composition of
147
iron (hydr)oxides depends on a number of other variables (such as pH and solution
composition) as well as the details of a complicated path through Eh-pH space during
electrolysis, it is extremely difficult to predict. It is possible that there is no significant
change in composition within certain ranges of current density, but large differences
between ranges. This could lead to pockets of stability within which, changing the
current density has no effect on arsenic removal capacity.
In the lower plot of Figure 3.13, current densities seem to follow a progres-
sion in arsenic removal capacity, with 0.02 mA/cm2 > 1.1 mA/cm2 > 5.0 mA/cm2 >
10 mA/cm2 ≈ 30 mA/cm2 ≈ 100 mA/cm2. However, the calculated removal capacity
for j = 5.0 mA/cm2 in Table 3.5 is very close (within 6%) to that of j = 10 mA/cm2.
The lower amount of arsenic removed for j = 5.0 mA/cm2 indicates that the apparent
difference seen in Figure 3.13 is due to a lower initial arsenic concentration rather than
an actual increase in removal capacity. Taking this into account, the removal capac-
ity in descending order is 0.02 mA/cm2 > 1.1 mA/cm2 > 5.0 mA/cm2 ≈ 10 mA/cm2
≈ 30 mA/cm2 ≈ 100 mA/cm2. From batch tests in SBGW-1, the removal capacity
in descending order is 0.07 mA/cm2 > 0.30 mA/cm2 ≈ 0.70 mA/cm2 ≈ 1.1 mA/cm2.
The removal capacity in SBGW-1 is not directly comparable to the removal capacity in
SBGW-2 because the differences in water composition have a direct affect on removal
capacity. However, it is possible to combine the data into an expected progression. From
comparisons of j = 1.1 mA/cm2 in SBGW-1 and SBGW-2, removal capacity decreases
in SBGW-2 relative to SBGW-1 (reasons for this effect are discussed more below). The
removal capacity of j = 0.07 macmsq in SBGW-1 is the same as that of j =0.02 mA/cm2
148
in SBGW-2 - this suggests that the removal capacity of 0.07 mA/cm2 would be less than
that of j =0.02 mA/cm2 if both were used on the same water. Similar comparisons
yield the following expected progression, now using ranges of current density to indicate
pockets of similar removal capacity: 0.02 mA/cm2 > 0.07 mA/cm2 > 0.30 - 1.1 mA/cm2
> 5.0 - 100 mA/cm2. These results are consistent with both Kumar et al. (2004) and
Gomes et al. (2007), who found no effect of current density on arsenic removal efficiency
within a range of j = 0.64 - 1.53 mA/cm2 and j = 3-30 mA/cm2 respectively. Both of
these ranges fall within the current density pockets of stable arsenic removal capacity10.
It has been proposed that changes in arsenic removal capacity with current
density are due to a change in iron (hydr)oxides composition. However, recall from the
Pourbaix diagram of iron (Figure 2.2) as well as Section 2.5.2 that as the cell poten-
tial (EH) increases, oxygen evolution becomes thermodynamically favorable. Once this
occurs, oxygen evolution will compete with iron dissolution for electrons, lowering the
current efficiency (Equation 2.21) and effectively producing fewer iron (hydr)oxides per
coulomb than possible at lower cell potentials. In this system, Eh is controlled by current
density, and thus oxygen evolution would compete for electrons more effectively as the
current density increases. Given that the removal capacity between j = 5.0 mA/cm2
and 100 mA/cm2 is so similar, oxygen evolution is unlikely, since the effect would be-
come progressively detrimental to arsenic removal with increasing current density. In
addition, no decrease in efficiency was seen by Chen et al. (2000) in the current density
range j = 1.25 - 10.89 mA/cm2 or Gomes et al. (2007) in the range j = 3 - 30 mA/cm2,10Note that Gomes et al. (2007) looked at arsenic removal in wastewater at pH 2.4. At this low pH, the
ranges of current density pockets are likely to be different from those at pH = 7, so it may be coincidencethat the tested range falls into a pocket of stable removal capacity.
149
which should have crossed the oxygen evolution threshold around j = 5 mA/cm2 if the
explanation above is correct. In the case of Gomes et al. (2007), this could be due to the
low pH used (pH = 2.4). The thermodynamic threshold of oxygen evolution increases
steeply as pH lowers, as demonstrated in Figure 2.2. Polarization curves in synthetic
Bangladesh water can detect oxygen evolution if it is occurring (see Section 3.3.10).
0 10 20 30 40 50
010
020
030
040
050
060
0
Current Processing Time (min)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
8060
4020
0
% A
rsen
ic R
emov
ed
1.1 mA cm2
0.7 mA cm2
0.3 mA cm2
0.07 mA cm2
No External VoltageBangladesh LimitWHO Limit
Figure 3.14: Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function current processing time for Matrix 1 batchexperiments for current densities j = 0.07, 0.30, 0.70, and 1.1 mA/cm2. Also includedare batch tests performed with no external voltage (blanks). Percent arsenic removed isbased on the average initial arsenic concentration. The WHO and Bangladesh maximumallowable arsenic limits are shown as dashed and dotted lines respectively. Error barsrepresent the variability between identical batch tests.
Figure 3.14 shows the resulting arsenic concentration for Matrix 1 batch tests
150
as a function of current processing time, tCP . Consistent with Chen et al. (2000); Kumar
et al. (2004); Gomes et al. (2007), an increase in current density decreases the current
processing time required to reduce arsenic arsenic. The estimated treatment time to
reduce arsenic to the WHO limit is calculated in Table 3.5, revealing a similar progression
for SBGW-1 tests. It is difficult to directly compare the treatment times listed for
SBGW-2 tests because different values of A/V were used to obtain the extreme current
densities with available equipment.
Because of the tradeoff between operating variables j, q, tCP , and A/V (Equa-
tion 2.34), tCP was not always constant for a given q dosage between current densities.
In fact, tCP varied from minutes to days between j = 100 mA/cm2 and j = 0.02 mA/cm2
(see Table 3.2). It is possible that some, or even all, of the increased arsenic removal
capacity seen at the lowest current density is due to the drastically increased contact
time. In an attempt to isolate the effect of current density, one batch test was run at j
= 1.1 mA/cm2, qtotal = 100 C/L, using the same values of tCP used in the Matrix 2 run
for j = 0.02 mA/cm2 (i.e. a slow- j = 1.1 mA/cm2 run compared to the relatively fast -
j = 1.1 mA/cm2 run from Matrix 2 ). Results for this run are shown in Figure 3.15 and
listed in Table 3.5 under other.
It is clear from Figure 3.15 that some, but not all, of the efficiency gains seen
at j = 0.02 mA/cm2 can be recouped by drastically increasing the current processing
time. The charge density required to reach the WHO limit for j = 1.1 mA/cm2 drops
from qWHO = 150 C/L to qWHO = 50 C/L with the additional time. This increase
in efficiency with additional processing time indicates that either the system is far from
151
0 25 50 75 100 125 150 175
010
030
050
0
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
8060
4020
0
% A
rsen
ic R
emov
ed
Slow Tcp −0.02 mA cm2, 19.4 hr totalSlow Tcp − 1.1 mA cm2, 19.4 hr totalFast Tcp −1.1 mA cm2, 1.2 hr totalBangladesh Limit
WHO Limit
0 25 50 75 100 125 150 175
050
100
200
300
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
9080
7060
50
% A
rsen
ic R
emov
ed
Slow Tcp −0.02 mA cm2, 19.4 hr totalSlow Tcp − 1.1 mA cm2, 19.4 hr totalFast Tcp −1.1 mA cm2, 1.2 hr totalBangladesh Limit
WHO Limit
Figure 3.15: Arsenic removal (concentration on left vertical axis; percent arsenic removedon right vertical axis) as a function of charge density, q, for three batch tests; j = 0.02mA/cm2 with tCP total = 19.4 hours, j = 1.1 mA/cm2 with the same tCP total = 19.4hours, and the original Matrix 2 batch test run for j = 1.1 mA/cm2 using the muchshorter tCP total = 1.2 hours. Lower plot presents same data as upper plot on a reducedvertical scale to show detail at low arsenic concentrations. Percent arsenic removed isbased on the average initial arsenic concentration. The WHO and Bangladesh maximumallowable arsenic limits are shown as dashed and dotted lines respectively. Error barsrepresent the variability between identical batch tests.
152
equilibrium when the batch test is performed quickly, or that oxidation of As[III] to As[V]
is much slower than the processing times used in the original batch test. In either case,
increasing the processing time can clearly be used to increase the efficiency of ECAR. It
is important to note, however, that j = 0.02 mA/cm2 still demonstrates a higher arsenic
removal capacity than j = 1.1 mA/cm2 even with processing time is kept constant. This
effect is indeed likely due to a significant change in the composition of iron (hydr)oxides
produced.
3.3.5 Effect of current processing time
0 15 30 45 60 75
0.0
0.2
0.4
0.6
0.8
1.0
Charge Density (C/L)
Nor
mal
ized
Ars
enic
Con
cent
ratio
n
100
8060
4020
0
% A
rsen
ic R
emov
ed
Slow Tcp − 19.4 hr totalFast Tcp − 7.5 hr total
Figure 3.16: Arsenic removal (normalized concentration on left vertical axis; percentarsenic removed on right vertical axis) as a function of charge density, q, for two batchtests at j = 0.02 mA/cm2 in which the only difference is a decrease in tCP by 60% forthe Fast test relative to the slow tests. Both batch tests used Matrix 2 procedures andSBGW-2 water with initial arsenic = 600 µg/L.
153
The current processing time, tCP , is generally used only to adjust the charge
density (via Equation 2.34). However, tCP may affect arsenic removal capacity in ECAR
in a number of different ways that are independent of the applied charge density. As
discussed in Section 2.5.2, increasing tCP may allow for increased coagulation relative to
mixing due to the presence of the electric field during tCP . It could also allow for in-
creased oxidation of As[III] to As[V], a process that does not occur, or occurs much more
slowly, during the mixing stage. To determine if any of these processes are significant in
ECAR, two batch tests were performed at j = 0.02 mA/cm2 in which the only change
was to decrease tCP by 60% for each value of charge density q (Figure 3.16). The faster
run lags in terms of arsenic removal until around q = 25 C/L, at which point it catches
up with the slower run. This suggests that there is a separate effect of tCP on arsenic
removal until some threshold is reached, at which point the effect disappears. This is
consistent with a limitation based on the oxidation of As[III] to As[V] - at some point,
oxidation will be complete and no longer delay arsenic removal. If the limitation were
due to coagulation, then the fast run would consistently show less arsenic removal.
These results, along with those presented in Figure 3.15 indicate that tCP can
affect removal capacity independent of charge density or current density. This is an
important finding that indicates (1) the arsenic removal capacity values in Table 3.5
must be taken in the context of their specific tCP values (which are based on the same
A/V ratio used in batch tests) and (2) tCP should be included as a controlling variable
along with j, q, and tM , that can be tuned to optimize performance metrics in a local
water environment. It is worthwhile to note that the differences in tCP between long
154
and slow tests is quite large - total treatment to q = 150 C/L took 19.4 hours in the
slow case presented in Figure 3.15, compared to 1.2 hours in the original batch test. The
slow case may not be viable in the field if treatment is to occur continuously, or at the
longest, overnight. It is not known if smaller changes in tCP significantly affect removal
capacity. Future tests are planned to determine the benefit available with smaller time
losses.
3.3.6 Effect of phosphate and silicate on arsenic removal
It is difficult to directly measure the effect that increased levels of phosphate
and silicate in the water have on arsenic removal using this dataset because the Matrix
1 and Matrix 2 procedures utilized different surface-area-to-volume (A/V) ratios, and
consequently different current processing times. In addition, the Matrix 1 experiments
used a separated cathode cell, which will speed up the net oxidation of As[III] relative
to a single chamber cell, since reduction of As[V] back to As[III] is no longer possible.
Thus only a general idea of the effect of silicate and phosphate on arsenic removal via
ECAR can be given.
Figure 3.17 shows the final arsenic concentration as a function of charge density,
q, after ECAR treatment of low phosphate, zero silicate synthetic Bangladesh groundwa-
ter (SBGW-1) and average phosphate, average silicate synthetic Bangladesh groundwater
(SBGW-2) at current density j = 1.1 mA/cm2. From the figure, a higher charge density
is required to achieve the WHO limit (10 µg/L) of arsenic in SBGW-2 water. For low
phosphate, zero silicate water, the first measured point with final arsenic less than 10
155
0 100 200 300 400
010
030
050
0
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
8060
4020
0
% A
rsen
ic R
emov
ed
Low Phosphate, Zero SilicateAve Phosphate, Ave SilicateBangladesh LimitWHO Limit
Figure 3.17: Arsenic removal (concentration on left vertical axis; percent arsenic removedon right vertical axis) as a function of charge density, q, for low phosphate, zero silicatesynthetic Bangladesh groundwater (SBGW-1, using Matrix 1 procedure) and averagephosphate, average silicate synthetic Bangladesh groundwater (SBGW-2, using Matrix2 procedure). ECAR treatment occured at j = 1.1 mA/cm2 in both cases. All datahas been normalized to have an initial arsenic concentration of 580 µg/L for comparisonpurposes. Lines connecting the data points have been added to guide the eye only. TheWHO and Bangladesh maximum allowable arsenic limits are shown as dashed and dottedlines respectively.
156
µg/L is q=85 C/L, whereas q=150 C/L is required in average phosphate, average silicate
water, an increase of 76%. From Faraday’s law (Equation 2.19), this corresponds also to
an increased dosage of Fe[III] or Fe[II] by 76%, or 1.76 times, as well. SBGW-2 water
contains 19.5 mg/L Si and 1.3 mg/L P (see Table 3.1), compared to 0 mg/L Si and 0.01
mg/L P in SBGW-1. The ratio of Si:Astot in SBGW-2 is ≈ 33:1 and that of P:Astot is
≈ 2:1.
Recall from Section 2.4.6 that Roberts et al. (2004) investigated the amount
of Fe[II] and Fe[III] required in coprecipitation experiments to reduce 500 ppb of As[III]
and As[V] to below 50 ppb with and without 30 mg/L Si (a Si:Astot ratio of 60:1) and 3
mg/L P (a P:Astot ratio of 6:1). Over 2.5 and 7 times the Fe[III] was needed for As[III]
and As[V] respectively in the presence of Si and P. About 1.5 and 5 times the Fe[II] was
needed for As[III] and As[V] respectively. If the increase in Si and P produced a linear
detrimental effect on arsenic removal, the Roberts et al. (2004) data would imply 1.25-
3.5 times the Fe[III] or 0.75-2.5 times the Fe[II] additions would be required to remove
arsenic from SBGW-2 water (which has approximately half the Si and P added per mg
As as the Roberts et al. (2004) data, along with containing half As[III] and half As[V]
initially). In this case, the increase in dosage required by ECAR in the presence of Si
and P is similar to that required by coprecipitation with Fe[II] or Fe[III]. In addition,
SBGW-2 has bicarbonate, which can act with Si to enhance the detrimental effect of P
(Section 2.4.6). We can nevertheless say that the detrimental effect on arsenic removal
via ECAR in the presence of Si and P is consistent with the detrimental effect seen in
arsenic removal via coprecipitation with Fe[II] or Fe[III].
157
3.3.7 Parameter trends, tradeoffs, and implications for Bangladesh
The ECAR parameters j, q, tCP , and tM were all found to have an effect on
arsenic removal capacity in synthetic Bangladesh groundwater at pH = 7. Removal ca-
pacity was found to decrease with current density excepting certain ranges of stability
defined by: 0.02 mA/cm2 > 0.07 mA/cm2 > 0.30 - 1.1 mA/cm2 > 5.0 - 100 mA/cm2.
Increasing the charge density decreases the current processing time if A/V is held con-
stant. Increasing the charge density requires either a higher operating current (and hence
a higher operating voltage, leading to increased energy consumption) or an increase in
the active electrode surface area.
Arsenic removal was found to increase with charge density, q. The charge
density required to reduce 550 - 580 µg/L arsenic to below the WHO limit with tM =
60 min was q = 25 - 180 C/L. The charge density requirement was found to be sensitive
to the composition of sample water and operating conditions such as j and tCP . It should
not be taken as an absolute value, but calibrated to the specific water environment and
conditions in use. In general, higher arsenic removal capacity implies a lower charge
density requirement to reach a given concentration limit. At a given current density,
increasing the charge density requires an increase in the current processing time.
Increasing the current processing time was found to increase the arsenic removal
capacity independently of q. Increasing the mixing time, tM was found to increase
arsenic removal capacity rapidly for the first 10 - 15 minutes, and very slowly thereafter
(data presented in Section 3.2). Thus there is a tradeoff between removal capacity
and treatment time for both tCP and tM . Filtration times greater than 3 days (in
158
sedimentation tests) were found to increase the arsenic removal capacity slightly at for
a large time tradeoff. This effect differs from an increase in tM only because of the lack
of stirring during filtration (i.e. increasing tM is likely to be more time efficient).
In batch tests, the highest arsenic removal capacity in SBGW-2 (including
relevant levels of bicarbonate, phosphate, and silicate to give a reasonable estimate of
performance in Bangladesh) with a final concentration at the WHO limit, was achieved at
j = 0.02 mA/cm2, q = 25 C/L, tCP = 416 min (∼ 7 hrs), tM = 60 min. Total treatment
time (8 hrs) was just above the target for a community water system (6-7 hrs). In light
of the long treatment time, a second effective set of conditions in SBGW-2 was j =
1.1 mA/cm2, q=150 C/L, tCP = 69 min, tM = 60 min. The total treatment time was
2.15 hrs. Both of these parameter sets have advantages (the first in terms of operating
costs, and the second in terms of treatment time) and choosing between them would
require more detailed knowledge on the importance of reduced operating costs compared
to treatment time. The actual operating conditions should also take into account relevant
safety factors.
3.3.8 Adsorption using post-synthesis ECAR-generated adsorbent
In order to explore the significance of producing iron (hydr)oxides in-situ with
arsenic during ECAR, experiments were performed using iron (hydr)oxides that were
not generated in-situ. Post-synthesis ECAR-generated iron (hydr)oxides adsorbent,
or PS-EGA adsorbent, was created by running the ECAR dosing process in synthetic
Bangladesh groundwater containing no arsenic. PS-EGA was placed in contact with ar-
159
senic after being freshly made (fresh-PS-EGA) as well as after being aged for 60 minutes
(aged-PS-EGA). Figure 3.18 shows the adsorption achieved by both fresh- and aged-PS-
EGA adsorbent as a function of contact time. It is clear that fresh-PS-EGA adsorbent
removes arsenic much more quickly than the aged-PS-EGA adsorbent, leading to a fi-
nal arsenic value that is 10 times lower after 120 minutes of contact time. In terms of
removal capability, fresh-PS-EGA adsorbent is capable of reducing arsenic to below the
Bangladesh limit after about 45 minutes of contact time, as well as approach the WHO
limit at 120 minutes (final arsenic concentration was 22 µg/L). However, aged-PS-EGA
adsorbent comes nowhere near even the Bangladesh limit after 120 minutes of contact
(final arsenic concentration was 190 µg/L). This implies that the age of PS-EGA adsor-
bent has a huge impact on the final arsenic concentration achievable using post-synthesis
ECAR-generated adsorbents. Even after aging for only 60 minutes, treatment of syn-
thetic Bangladesh groundwater to below the Bangladesh legal limit becomes impossible
without (perhaps) a large increase in adsorbent added or (possibly) a large increase in
contact time. Since 60 minutes is too short to transport pre-formed adsorbent in a useful
way, in-situ generation is necessary to achieve the arsenic removal capacity and arsenic
removal capability of ECAR.
It appears from Figure 3.18 that the arsenic concentration for aged-PS-EGA
plateaus after about 80 minutes of contact, while the arsenic concentration for fresh-
PS-EGA continues to decrease up to 120 minutes of contact. Figure 3.19 and Table 3.6
show results after 1-2 days of quiescent contact time following the mixing period (in each
case, samples were filtered using 0.1µm membranes before arsenic analysis). From this
160
0 20 40 60 80 100 120
010
020
030
040
050
060
0
Contact Time (min)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
8060
4020
0
% A
rsen
ic R
emov
ed
PSA−FreshPSA−Aged 60 minBangladesh LimitWHO Limit
Figure 3.18: Arsenic removal (concentration on left vertical axis; percent arsenic re-moved on right vertical axis) as a function of contact time for PSA experiments usingpost-synthesis ECAR-generated adsorbents, both freshly prepared (PSA-Fresh) and agedfor 60 minutes (PSA-Aged 60 min). Adsorbent was generated using ECAR operatingconditions j = 5.0 mA/cm2, q = 175 C/L, tCP = 17.27 min in arsenic-free SBGW-2water. Note that the initial arsenic level is estimated based on the spiked amount (seeSection 3.2.8). Percent arsenic removed is based on the average initial arsenic concentra-tion. The WHO and Bangladesh maximum allowable arsenic limits are shown as dashedand dotted lines respectively. Error bars represent the variability between identical batchtests.
161
0.5 1.0 1.5 2.0
050
100
150
200
250
Quiescent Contact Time (Days)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
9080
7060
% A
rsen
ic R
emov
ed
PSA−FreshPSA−Aged 60 minECAR TreatmentBangladesh LimitWHO Limit
Figure 3.19: Arsenic removal (concentration on left vertical axis; percent arsenic removedon right vertical axis) as a function of quiescent contact time for PSA experiments usingpost-synthesis ECAR-generated adsorbents, both freshly prepared (PSA-Fresh) and agedfor 60 minutes (PSA-Aged 60 min), and compared to ECAR treatment at the samecurrent density (j=5.0 mA/cm2) and a similar current processing time (tCP = 16.33min) as that used to generate the PSA adsorbents. This is a continuation of Figure 3.18(note however the change in x-axis units). In all cases, quiescent contact began after aperiod of active mixing (tM = 60 min in the case of ECAR treatment, and tM = 120minutes in the case of PSA experiments). Initial arsenic concentration in all three caseswas Astot = 590 ± 40 µg/L. Percent arsenic removed is based on the average initialarsenic concentration. The WHO and Bangladesh maximum allowable arsenic limitsare shown as dashed and dotted lines respectively. Error bars represent the variabilitybetween identical batch tests.
162
Table 3.6: Comparison of arsenic removal capability and arsenic removal capacity usingECAR treatment versus treatment with freshly made (PSA-fresh) and aged 60 min(PSA-aged) post-synthesis ECAR-generated adsorbents. Arsenic removal capacity hasbeen calculated assuming that all of the applied charge (during ECAR treatment orECAR-generation of adsorbent) is used to form Fe 3+ ions in solution.
Residual Residual Residual Residual As Removal As RemovalInitial As at As at As at As at Capacitya at Capacitya At
Treatment As 60 min 120 min 1 day settle 2 day settle 60 min 2 day settle(µg/L) (µg/L) (µg/L) (µg/L) (µg/L) (µg-Asrem/ (µg-Asrem/
mg-Fe 3+) mg-Fe 3+)
ECARb 540 10 NA 12 NA 18.3 18.3 c
PSA-Fresh 610d 37 22 23 22 17.0 17.4PSA-Aged 610d 220 190 110 100 11.5 15.1
a Arsenic removal capacity (in µg-As-removed/mg-Fe 3+) assumes that all charge added is used to generateFe 3+ ions in solution.
b ECAR treatment used similar operating parameters (j = 5.0 mA/cm2, q = 175 C/L, tCP = 16.33 min) asthose used to generate PS-EGA adsorbents (j = 5.0 mA/cm2, q = 175 C/L, tCP = 17.27 min) and usedthe same sample water composition (SBGW-2). After 60 minutes of mixing, the ECAR sample was left toquiescent settling, and thus no data is available at 120 minutes of mixing. In addition, data was only takenafter 1 day of quiescent settling.
c The arsenic removal capacity here is assumed to be the same as it was after 1 day of quiescent settling, basedon showing no change in arsenic between 60 minutes and 1 day of settling (and assuming no change wouldoccur between 1-2 days).
d Initial arsenic values are estimated from spiked amount (see Section 3.2.8 for explanation).
163
figure, it is clear that the arsenic concentration for fresh-PS-EGA has reached a plateau,
while the arsenic concentration for aged-PS-EGA continues to fall. This is similar to
the trend seen in adsorption via coprecipitation compared to PSA. Recall from Section
2.4.5 that adsorption onto pre-formed iron (hydr)oxides (PSA) tends to take longer
than adsorption via coprecipitation because aggregates have already formed, and ions
must diffuse into the aggregates before steady-state can be achieved. The formation of
aggregates is likely the cause of the slower arsenic removal and lower arsenic removal
capacity seen in aged-PS-EGA compared to fresh-PS-EGA.
Table 3.6 lists the arsenic concentration and inferred arsenic removal capacity
(assuming all charge is used to generate Fe 3+ ions) for the PS-EGA adsorbents as well
as ECAR treatment. In this comparison, the ECAR treatment used similar operating
parameters (j = 5.0 mA/cm2, q = 175 C/L, tCP = 16.33 min) as those used to generate
PS-EGA adsorbents (j = 5.0 mA/cm2, q = 175 C/L, tCP = 17.27 min) and treated the
same sample water composition (SBGW-2). Therefore, differences in arsenic removal
capacity are not likely due to differences in the iron (hydr)oxides composition of the
adsorbent produced or composition differences in the sample. From Table 3.6, one can see
that the initial arsenic concentration is a bit lower for the ECAR treatment test compared
to the PS-EGA tests, however this difference is almost within measurement error (10%)
and is automatically normalized during the arsenic removal capacity calculation. One can
see from Table 3.6, as well as Figure 3.19, that ECAR treatment performs better than PS-
EGA adsorbents both in the short term (after the standard 60 minutes of mixing contact)
and longer term (after 1-2 days of quiescent contact). Notably however, the performance
164
of fresh-PS-EGA is not far behind that of ECAR in terms of arsenic removal capability
(reducing to 22 µg/L as opposed to 10 µg/L) or arsenic removal capacity (17.0 compared
to 18.3 µg-Asrem/mg-Fe 3+). This indicates that only a small relative advantage is gained
from arsenic adsorption that occurs during the current processing time of ECAR. This
is interesting to note because one process that may occur during electrolysis is enhanced
coagulation of forming iron (hydr)oxides particles due to the applied electric field (see
Section 2.5.2). Enhanced coagulation would tend to age the generated iron (hydr)oxides
(in terms of aggregate formation) at an accelerated rate compared to simple mixing.
From the comparison of fresh-PS-EGA to 60 min aged-PS-EGA, a small acceleration in
aging could significantly reduce the arsenic removal capacity of the generated adsorbent.
This does not seem to be the case, indicating that the iron (hydr)oxides formed during
ECAR are not significantly more aggregated due to electrolysis.
It is also interesting to note that the final arsenic concentration in fresh-PS-
EGA solution does not approach the final concentration due to ECAR treatment, even
after 2 days of quiescent contact (Figure 3.19). If the only difference between the two
adsorbents were aggregation, the one would expect the final concentration to converge
as more time was allowed for diffusion. This indicates that some advantage of in-situ
generation during ECAR cannot be made up for using pre-formed adsorbent.
3.3.9 As[III] removal in ECAR
It is clear from Matrix 1 and Matrix 2 results (Figures 3.12 and 3.13) that
ECAR is capable of reducing both As[III] and As[V] from 300 µg/L to levels below the
165
0 50 100 150
010
020
030
0
Charge Density (C/L)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
100
8060
4020
0
% A
rsen
ic R
emov
ed
AsIIIAsVBangladesh LimitWHO Limit
Figure 3.20: Post-treatment As[III] and As[V] (concentration on left vertical axis; percentarsenic removed on right vertical axis) as a function of charge density, q. This batch testwas run on SBGW-2 water with operating conditions j = 1.1 mA/cm2, tCP = 66.58 min,Matrix 2 procedure. These results were typical of all current densities in Matrix 2 (i.e.0.02, 1.1, 5.0, 10, 30, and 100 mA/cm2). Lines have been added to guide the eye only.Percent arsenic removed is based on the average initial arsenic concentration. The WHOand Bangladesh maximum allowable arsenic limits are shown as dashed and dotted linesrespectively. Error bars represent the variability between identical batch tests.
166
NA0
20
40
60
80
100
120
140
160
ECAR PSA-Fresh PSA-Aged
AsI
II C
on
cen
trati
on
(u
g/
L)
(In
itia
l A
sIII
= 3
00
+/
- 2
0 u
g/
L)
After 60 min Mixing
After 23 hr settle
Figure 3.21: Comparison of the final As[III] concentration (initial concentration 300µg/L) for SBGW-2 water after ECAR treatment to adsorption with freshly madepost-synthesis ECAR-generated adsorbent (PSA-Fresh) and aged post-synthesis ECAR-generated adsorbent (PSA-Aged). ECAR treatment was performed under similar condi-tions to the generation of post-synthesis adsorbent. See Section 3.3.8 and Section 3.2.8for experimental conditions.
167
WHO limit of 10 µg/L at pH 7. Kumar et al. (2004) has attributed similar behavior
(using EC with iron) to net oxidation of As[III]. Net electrochemical oxidation of As[III]
to As[V] is a significant side-reaction that may occur during the electrochemical process in
ECAR. Oxidation of As[III] may also occur in parallel with Fe[II] oxidation by dissolved
O2 (see Section 2.4.3). Oxidation of As[III] is advantageous if ECAR-generated iron
(hydr)oxides have a higher binding affinity for As[V] than As[III]. Recall from Section
2.4.4 that coprecipitation with alum, ferric chloride (Fe 3+), and ferric sulfate (Fe 2+) is
less efficient for As[III] removal than As[V] under comparable conditions (Hering et al.,
1996; Gulledge & Oconnor, 1973; Sorg & Logsdon, 1978; Roberts et al., 2004; Meng et al.,
2002; Shen, 1973; Leckie et al., 1980; Meng et al., 2000). However, experiments on pre-
formed iron (hydr)oxides, including ferrihydrite and goethite, have been discrepant at
arsenic concentrations near those used here (8 µM). A crossover pH (at which the binding
strength of As[III] and As[V] are equal) of 7.0 has been reported using ferrihydrite at
concentrations of 13.3 µM arsenic (Pierce & Moore, 1982). Yet, Dixit & Hering (2003)
report a crossover pH of 8.5 for ferrihydrite at similar arsenic concentrations (see Section
2.4.4 for more discussion). In light of the disagreements in the literature, it is not clear if
the ability to reduce As[III] concentrations comes primarily from a high affinity of ECAR-
generated iron (hydr)oxides for As[III] at pH 7 or from the net oxidation of As[III] and
subsequent adsorption as As[V].
Figure 3.20 shows the As[III] and As[V] concentrations as a function of charge
density during one of the Matrix 2 batch tests (SBGW-2 water, j = 1.1 mA/cm2, tM
= 60 min, q and tCP varied). Despite starting from similar initial concentrations, As[V]
168
is reduced to less than 10 µg/L after only 50 C/L passed, while As[III] requires up to
150 C/L11. Even as a lower bound, this implies that As[V] is removed more efficiently
than As[III] (i.e. requires less charge per mg arsenic removed). Since the composition of
iron (hydr)oxides produced during ECAR are a function of the current density used, the
results shown would only apply to j = 1.1 mA/cm2. However, these results were typical
of all measured current densities, including 0.02, 1.1, 5.0, 10, 30, and 100 mA/cm2,
implying that As[V] is more efficiently removed within the tested current density range.
This is consistent with a net oxidation of As[III]. It is also consistent with a slightly
lower binding affinity for As[III] , requiring more time, as opposed to more charge, to
adsorb to iron (hydr)oxides.
One way to separate electrochemical oxidation of As[III] from adsorption of
As[III] and/or oxidation of As[III] via parallel Fe[II] oxidation is to look at the As[III]
concentration during post-synthesis ECAR-generated (PS-EGA) adsorbent experiments
(Section 3.3.8). In these experiments, As[III] is not present during electrolysis, and so
cannot be oxidized by the anode12. It can, however, adsorb to the PS-EGA adsorbent as
As[III] or oxidize in parallel with Fe[II]. Figure 3.21 compares the As[III] concentration
after ECAR treatment (including 60 minutes of mixing) to adsorption onto PS-EGA ad-
sorbent after 60 minutes of contact and 23 hours of quiescent contact. ECAR treatment
has been chosen to closely match the conditions of adsorbent generation for PS-EGA11Note that these numbers could only be used to calculate a lower bound on the arsenic removal
capacity for As[V] and As[III], since the net As[V] concentration may grow during electrolysis if netoxidation of As[III] is occurring. It is also possible that the As[V] concentration could lower if netreduction to As[III] is occurring, though this is unlikely.
12Oxidation of As[III] could occur without electrolysis, as measured in synthetic Bangladesh ground-water batches during storage (see Section 3.2.3). However, this effect was only significant after 2-3 daysof storage.
169
(see Section 3.3.8). It is clear that less As[III] remains after ECAR treatment than after
adsorption with PS-EGA. However, 90% and 40% of the As[III] is adsorbed by fresh-PS-
EGA and aged-PS-EGA respectively within 60 minutes of contact. This clearly indicates
that ECAR-generated adsorbent has some affinity for As[III], even in the presence of com-
peting As[V] as well as background ions present in SBGW-2 water. The As[III] continues
to adsorb over 23 hours of quiescent contact. Thus one could argue that the lower levels
of As[III] seen during ECAR treatment are entirely due to the increased contact time
allowed by arsenic contact during electrolysis (i.e. in the ECAR case, arsenic was in con-
tact with ECAR-generated iron (hydr)oxides for 16 minutes longer during electrolysis).
However, even after 23 hours of contact, As[III] levels in the PS-EGA adsorbent have
not reached those of ECAR treatment after 76 minutes of contact, indicating that the
extra contact time is not the only advantage. This provides evidence that some elec-
trochemical oxidation of As[III] is occurring during electrolysis. Oxidation may explain
part or all of the advantage of in-situ generation over post-synthesis adsorption see in
Section 3.3.8.
3.3.10 Polarization studies
Figure 3.22 shows an anodic polarization scan for iron in plain perchlorate
solution (0.1M KClO4) compared to perchlorate plus synthetic Bangladesh groundwater
(0.1M KClO4 + SBGW) at a scan rate of ν = 0.1 mV/s. The slow scan rate allows the
current to come close to equilibrium at each potential step, allowing for some idea of the
equilibrium current density that corresponds to each voltage in SBGW (see discussion in
170
−0.7 −0.5 −0.3 −0.1 0.1
Potential (V) vs SHE
Cu
rre
nt D
en
sity ((
mA
cm
2))
1e−
04
0.0
10.1
110
1.1 mA cm2
0.7 mA cm2
0.3 mA cm2
0.07 mA cm2
0.02 mA cm2
5.0 mA cm2
10.0 mA cm2
30.0 mA cm
H AsO3 3H AsO2 4
-
HAsO42-
0.1 M KClO4
0.1 M KClO + SBGW4
10 M-8
1MFe/Fe
2+Fe /Fe(OH)3
2+1M
Fe Fe(OH)2
Fe(OH) Fe(OH)2 3
pH = 9
pH = 7
H AsO3 3 HAsO42-
0.0
01
Figure 3.22: Anodic polarization curve for iron wire electrode in 0.1M KClO4 (thin line)and 0.1M KClO4 + SBGW (thick line) scanned at ν = 0.1 mV/s. Horizontal lines indi-cate the current densities used in Matrix 1 (dotted) and Matrix 2 (dashed) experiments.Thermodynamic equilibrium potentials at 25C for appropriate redox couples are shownfor pH =7 (in red) - expected of natural groundwater - and pH = 9 (in blue) - close to thepH of the actual scan. A pourbaix diagram composed by Bang et al. (2005a) assumingsolid Fe, Fe(OH)2 and Fe(OH)3 exist was consulted to select appropriate lines at eachpH. If concentration appears in the equilibrium calculation, a range of concentrationsfrom 10−8M to 1M are shown.
171
Section 3.2.9). This in turn gives an idea of the interface potential expected in SBGW
at a given constant current density. Horizontal lines depict the current densities used in
Matrix 1 and Matrix 2 experiments. All scans went fully negative (to -1.4V vs SHE)
before the oxidation scan to eliminate any prior oxide layer buildup on the electrode.
A comparison of the scan in plain perchlorate to the scan in perchlorate plus
SBGW should reveal electrochemical effects due only to ionic species in the synthetic
groundwater. In Figure 3.22, the SBGW polarization scan is shifted anodically by ap-
proximately 50 mV compared to the plain perchlorate scan. This shift translates into
a large drop in current, especially at potentials < -0.1V, suggesting that species in the
SBGW are interfering with the dissolution processes on the iron electrode. The drop in
current could indicate slower kinetics, a different oxidation product distribution, and/or
a compact passive layer inhibiting but not preventing current flow. The shift is too large
to be attributed to pH changes alone (pHf = 9.14 and pHf = 9.60 in plain perchlorate
and perchlorate plus SBGW respectively). 0.1M perchlorate at was added to both solu-
tions, and thus the decrease in current is not due to the high resistivity of SBGW. Thus,
in addition to the overpotential required to overcome the high resistivity of Bangladesh
groundwater, some overpotential will be required overcome the drop in current measured
above, further increasing the energy consumption of ECAR treatment. The drop in cur-
rent relative to plain perchlorate decreases with increasing current density, indicating
that this effect is smaller at higher current densities (j > 5 mA/cm2). Thus higher cur-
rent densities may have some advantage over lower current densities in terms of energy
172
consumption13.
Notably, no iron passivation is seen within the range of the scan (i.e. up to
10 mA/cm2). Passivation of iron occurs when an oxide layer builds up on the electrode
surface, preventing current from passing. The onset is characterized by a steep decrease
in current during oxidation, falling as low as 30 - 50 µA/cm2 in perchlorate solution
(Jovanovic & Hackerman, 1998). The passivation potential of iron in perchlorate solution
is typically 0.544 ±0.030 V vs SHE (Jovanovic & Hackerman, 1998), well above the region
corresponding to j = 10 mA/cm2 in Figure 3.22. Unfortunately, the scan did not reach
current densities as high as 30 or 100 mA/cm2, though the fact that significant arsenic
was removed at these current densities through ECAR suggests that passivation in this
region did not occur. It is worthwhile to note that the pH of the solution during the
scan was likely as high as pH = 9.14 (see Section 3.2.9), and the passivation potential
is known to shift with pH. However, the passivation potential shifts cathodically (lower
in voltage) with increasing pH, making it more likely that passivation would be seen in
this system than in the real system at neutral pH.
The lack of passivation up to current density 10 mA/cm2 is significant. Mameri
et al. (1998) point out that all teams working with aluminum monopolar electrodes
to remove fluoride via electrocoagulation met with passivation problems. The lack of
passivation in iron over the current density region that is most effective for arsenic
removal is promising in terms of maintenance requirements and the ability to operate in
monopolar mode.13Note that higher current densities may also have energy consumption disadvantages if a higher
operating current is used (as opposed to decreasing the active electrode area), requiring a higher voltage.
173
It is also significant to see that the potential at current densities as high as 10
mA/cm2 is well below the potential required for oxygen evolution, which occurs near
0.8 V vs SHE at atmospheric pressure and pH = 7, and near 0.7 V vs SHE near pH =
9 (Misawa, 1973) . Oxygen evolution can compete with iron dissolution for electrons,
reducing the current efficiency (Equation 2.21). This leads to fewer iron (hydr)oxides
produced per coulomb and hence less arsenic removal capacity. Recall that one possible
explanation for the decrease in arsenic removal capacity seen at j > 5.0 mA/cm2 was the
onset of oxygen evolution. While the data in Figure 3.22 cannot rule out oxygen evolution
at j > 30 mA/cm2, it makes this explanation highly unlikely for j ≤ 10.0 mA/cm2. The
onset of oxygen evolution also limits the highest current density that could potentially be
effective in ECAR. Because that limit has not yet been reached, higher current densities
should be explored to look for possible increases in arsenic removal capacity (if the
composition of EGA changes favorably) that may occur before reaching the oxygen
evolution potential. Since higher current densities can deliver more charge in less time,
this could potentially provide a very favorable parameter space for arsenic removal.
Thermodynamic equilibrium potentials at 25C for appropriate redox couples
are shown in Figure 3.22 for pH =7 (in red) - expected of natural groundwater - and pH
= 9 (in blue) - close to the final pH of the scan. A Pourbaix diagram composed by Bang
et al. (2005a) assuming solid Fe, Fe(OH)2 and Fe(OH)3 exist was consulted to select
appropriate lines at each pH. If concentration appears in the equilibrium calculation,
a range of concentrations from 10−8M - 1M are shown. It is difficult to make direct
comparisons to ECAR treatment due to the high pH of the scan, however some general
174
comparisons are useful to suggest further testing.
Current densities j = 0.02 mA/cm2 and j = 0.07 mA/cm2 both occur to the
anodic side of the transition between Fe(0) and Fe[II] species and to the cathodic side of
the transition between Fe[II] to Fe[III] species for both pH 7 and 9. This suggests that the
iron (hydr)oxides produced at these current densities are based predominantly on Fe[II]
species rather than Fe[III], though there is some uncertainty due to the high pH during
the scan (pH = 9.14) relative to natural groundwater. Higher current densities fall to the
anodic side of the transition between Fe[II] and Fe[III] species, making it more likely that
the iron (hydr)oxides are composed of Fe[III]. The increased arsenic removal capacity
seen at current densities j = 0.02 and 0.07 mA/cm2 suggests that Fe[II] species may
have higher sorption capacities than Fe[III] species. Alternatively, Fe[III](hydr)oxides
formed from the oxidation of Fe[II] may have a higher adsorption capacity for arsenic
than Fe[III](hydr)oxides formed from Fe[III] directly. Roberts et al. (2004) found that
both As[III] and As[V] removal efficiencies were higher with the chemical addition of
Fe[II] salts (quickly oxidized to Fe[III]) than the addition of Fe[III] salts in simulated
groundwater containing phosphate and silicate. This suggests that Fe[III](hydr)oxides
produced from Fe[II] oxidation may have higher adsorption capacities than Fe[III] in
Bangladesh groundwater.
Interestingly, the equilibrium potential for As[III] oxidation to As[V] (H3AsO3/HAsO 2 –4
or H2AsO –4 ) is above the potential associated with current densities j < 0.7 mA/cm2 for
pH 7 and j < 5.0 mA/cm2 for pH 9. This means that it is thermodynamically favorable
for As[V] to reduce to As[III] at many current densities. Matrix 1 and Matrix 2 results
175
indicate that As[III] is effectively removed at all current densities tested - previously
attributed to As[III] oxidation followed by As[V] complexation to iron (hydr)oxides.
These results indicate the possibility that As[III] is effectively removed due to direct
complexation with iron (hydr)oxides rather than oxidation followed by complexation.
Due to kinetics and other effects, equilibrium potentials do not prove that As[V] is being
reduced at low current densities, but it does suggest that further testing would be appro-
priate. Note also that direct complexation of As[III] would be consistent with the results
showing significant As[III] removal using post-synthesis ECAR-generated adsorbent.
3.3.11 Sedimentation versus 0.1µm vacuum filtration
The arsenic removal capability of ECAR using 0.1µm filtration was compared
to that of ECAR using sedimentation (i.e. quiescent settling followed by decantation)
during both the Matrix 1 and Matrix 2 experiments. Figure 3.23 shows the the final
arsenic concentration of four Matrix 1 experiments utilizing 0.1µm filtration compared
to the experiments of the same operating conditions utilizing sedimentation. The four
experiments cover a range of settling times, Ts, before decantation (Ts = 1.9 - 6 days)
and represent experiments from a range of operating conditions (j = 0.07 - 1.1 mA/cm2,
q = 15 - 85 C/L, tCP = 3 - 50 min; see Figure 3.23 caption for exact conditions of each).
For the first three experiments (1-3), with Ts ≤ 3.1 days, the arsenic concentration in
settled samples is much higher (by 10 - 64 µg/L), that the 0.1µm filtered samples. In
one case (experiment 2), the final arsenic concentration of the 0.1µm filtered sample
is lower than the WHO limit of 10 µg/L, while the sedimentation sample (Ts = 1.9
176
0
20
40
60
80
100
120
140
160
(1) Ts=1.9 days (2) Ts=1.9 days (3) Ts=3.1 days (4) Ts=3.8 days,6 days
Fin
al A
rsen
ic C
on
cen
trati
on
(p
pb
) 0.1 um filtered (no settling)Settled and DecantedSettled Longer and Decanted
Figure 3.23: Comparison of the final arsenic concentration for 0.1µm filtered and sedi-mentation samples from the Matrix 1 series of experiments. Ts, listed below each seriesof bars, is the quiescent settling time allowed before decantation for the settled sample(0.1µm filtered samples were all filtered immediately after mixing). In the case of series4 (right), a second sample was removed and tested from the sedimentation beaker at alater time. The ECAR operating conditions for the above experiments were (1) j = 1.1mA/cm2, q = 23 C/L, tCP = 3 min; (2) j = 1.1 mA/cm2, q = 85 C/L, tCP = 5 min; (3) j= 0.07 mA/cm2, q = 25 C/L, tCP = 50 min; and (4) j = 0.70 mA/cm2, q = 15 C/L, tCP= 3 min. Each series has been normalized to its average initial arsenic concentration toavoid small deviations due to different starting arsenic levels - the average initial arsenicconcentration was 610 ± 40 µg/L.
177
0 5 10 15 20 25 30 35 40 45
050
100
150
200
250
300
Settling Time (Hr)
Ars
enic
Diff
eren
ce::
Set
tled
−− F
ilter
ed ((µµ
gL))
30 min mix before settling60 min mix before settling
Figure 3.24: The difference in arsenic concentration between samples filtered through a0.1µm membrane and samples decanted from the top of a beaker after ECAR treatment(j = 1.1. mA/cm2, q = 100 C/L) following a period of quiescent settling as a functionof the quiescent settling time. Two tests are shown in which the solution was mixed for30 minutes before quiescent settling (solid line, solid symbols) and mixed for 60 minutesbefore quiescent settling (dashed line, open symbols). Lines have been added to guidethe eye only.
178
days) is higher than even the Bangladesh limit of 50 µg/L. All of this indicates that
3.1 days is too short for sedimentation to be as effective or nearly as effective as 0.1µm
filtration, specifically with respect to reducing levels to below the WHO limit. For the
final experiment (4), the first sedimentation sample was decanted after settling for 3.8
days. This sample shows a lower arsenic concentration (by about 20 ppb) than the
immediately filtered sample, probably due to the additional complexation time allowed
by settling and/or increase oxidation of As[III] into As[V] due to atmospheric exposure
over time. An additional observation after 6 days shows the same arsenic concentration
as found after 3.8 days, indicating that the concentration has stabilized some time before
3.8 days. Thus arsenic removal from SBGW-1 water using ECAR with sedimentation
can be as good as or even better than (due to the increased complexation time) ECAR
with 0.1µm filtering, as long as the sample is allowed to settle for at least 3.8 days (or
possibly some time between 3.1 - 3.8 days).
At one point, salvaged hard drive magnets were attached to the bottom of
a beaker containing dosed and mixed SBGW-1 water (Matrix 1 procedure) and the
settling rate compared to a beaker containing the same effluent with no magnets. After
21 hours of settling, both beakers appeared clear on top with an orange precipitate
uniformly spread over the bottom of the beaker. There was a very slight outline of
the magnets made by precipitate, indicating that the ECAR-generated particles are
only slightly magnetic. QuickTest measurements of the supernatant showed the same
arsenic concentration within error (50 ±25 µg/L) for magnet-enhanced and non-magnet-
enhanced settling. A repeat of this test yielded the same results after 3.5 hours of settling,
179
though in this case, no visible outline of the magnet was observed in the precipitate. It
is likely that the ECAR-generated particles in this current density range (0.07 - 1.1
mA/cm2) are not magnetic enough for settling to be aided by low-cost magnets.
Using the Matrix 2 procedure and SBGW-2 water, sedimentation has been
compared to 0.1µm filtration over a period of two days by sampling the top effluent of a
quiescent sample after ECAR treatment (at j = 1.1 mA/cm2, q = 100 C/L, tCP = 44.7
min) and comparing the unfiltered effluent to the filtered effluent. Two sedimentation
samples were compared in order to see if sedimentation could be accelerated by initial
mixing, including one experiment with dosed water that was mixed for 30 minutes before
settling and another that was mixed for 60 minutes before settling. The results of both
are shown in Figure 3.24.
From Figure 3.24 it is clear that sedimentation could not replace 0.1µm vacuum
filtration immediately after mixing, at which time the sedimentation sample contains
275 µg/L more arsenic than the 0.1µm filtered sample. In fact, the difference in arsenic
concentration does not approach zero until nearly 45 hours, or 1.9 days, have passed (the
final difference in arsenic measured was 3 µg/L in both experiments). This implies that
a two-day storage time would be required to replace 0.1µm filtration with sedimentation
to yield the same results.
The sample that had been mixed for 60 minutes prior to settling did settle
slightly faster than the sample that had been mixed for 30 minutes, though the gap
began to close with time. Both sedimentation samples were 3 µg/L above the 0.1µm
filtered samples after 45 hours. This is consistent with a slowing of the settling rate of the
180
60-min sample relative to the 30-min sample over time. It is also consistent with reaching
a maximum level of removal via sedimentation. For example some particles may be too
small to settle but large enough to be removed on a 0.1µm membrane. In this case, the
60-min sample may have reached this limit anytime between 28 and 45 hours. Thus it
is possible that the additional mixing could reduce the required settling time by up to
about 10 hours, making the total required settling time as low as 1.5 days. Additional
testing with more frequent samples between 28 and 45 hours and longer mixing times
could determine if increasing the mixing time further does in fact reduce the settling
time, and if so, if there is a mixing time such that the settling time could be reduced
substantially.
3.4 Chapter Summary
Two bench-top electrochemical cells have been designed and incorporated into
a small-scale ECAR batch treatment procedure. Two recipes for synthetic Bangladesh
groundwater (SBGW-1 and SBGW-2) have been developed, differing in the levels of com-
petitive co-occuring solutes phosphate, silicate, bicarbonate, and sulfate. The resistivity
of SBGW-1 was measured to be ρe = 355Ω·cm using impedance spectroscopy (EIS).
This high resistivity will likely lead to a high voltage requirement in real Bangladesh
groundwater, potentially mitigated by the addition of a small amount of salt before
treatment.
The arsenic removal capability of ECAR in synthetic Bangladesh groundwater
(SBGW) at pH = 7 was demonstrated in two distinct groundwater recipes (SBGW-1 and
181
SBGW-2, composition in Table 3.1) by reducing initial concentrations of 550-580 µg/L
arsenic (including both As[III] and As[V] in ratio As[III]:As[V] = 1:1) to below the
WHO limit of 10 µg/L. This was explicitly demonstrated for current densities j = 0.30
- 1.1 mA/cm2 in SBGW-1 and j = 0.02 - 5.0 mA/cm2 in SBGW-2. Based on trends,
it is likely that the WHO limit would be reached for j = 0.07 mA/cm2 in SBGW-1
and j = 10, 30, and 100 mA/cm2 in SBGW-2 with a small increase in charge density.
The legal Bangladesh limit (50 µg/L) was explicitly reached in every case. Arsenic
removal capability in SBGW-2 is better indicator of removal capability in Bangladesh
than SBGW-1 due to the average levels co-occurring solutes known to compete with
arsenic for adsorption, such as phosphate and silicate.
The charge density required to reach the Bangladesh and WHO limit in batch
tests (extrapolated in cases where it was not explicitly reached) is listed in Table 3.5.
Values based on the WHO limit were used to calculate arsenic removal capacity, iron
capacity (assuming dissolution of Fe 3+), and total treatment time. The charge density
required to reach the WHO limit varied from qWHO = 25 - 70 C/L in SBGW-1 and
qWHO = 25-180 C/L in SBGW-2 depending on current density. The corresponding
arsenic removal capacity was 8 - 22 µg-Asrem/C in SBGW-1 and 3 - 22 µg-Asrem/C in
SBGW-2.
Arsenic removal capacity was found to be approximately constant within certain
ranges of current density, and highly variable between ranges. In order of decreasing
removal capacity, the pattern was: 0.02 mA/cm2 > 0.07 mA/cm2 > 0.30 - 1.1 mA/cm2
> 5.0 - 100 mA/cm2. These results are consistent with both Kumar et al. (2004) and
182
Gomes et al. (2007), who found no effect of current density on arsenic removal efficiency
within a range of j = 0.64 - 1.53 mA/cm2 and j = 3-30 mA/cm2 respectively. Both of
these ranges fall within the current density pockets of stable arsenic removal capacity.
Batch test runs Matrix 1 and Matrix 2 (described in Sections 3.2.6 and 3.2.7)
used charge density, q, and current density, j, as independent variables. The current
processing time, tCP (which is a function of q, j, and the electrode surface area to volume
ratio, A/V according to Equation 2.34), was allowed to vary over an order of magnitude
between tests at 1.1 mA/cm2 and 0.02 mA/cm2. Surprisingly, a 6-fold increase in removal
capacity was observed at 0.02 mA/cm2 compared to 1.1 mA/cm2. In order to determine
if this increase was due to j alone or the large difference in tCP , an additional batch
test was performed at 1.1 mA/cm2 using the same tCP as the 0.02 mA/cm2 batch test.
This revealed that most, but not all, of the increase in removal capacity originally seen
at 0.02 mA/cm2 was due to the increase in tCP . Comparisons of the two batch tests
at 1.1 mA/cm2 showed that removal capacity could be increased 3-fold at the cost of
increasing tCP from ∼ 1 hour to ∼ 19 hours. This finding indicates that increasing tCP
can increase arsenic removal capacity independently of q. This effect could be due to
the increase in contact time between EGA, the increase in duration of As[III] oxidation
to As[V], increased duration of coagulation in the presence of an electric field, or some
unknown effect. If it is due to the increase in contact time, then adjusting the mixing
time, tM should have the same effect as increasing tCP . Further testing is needed to
determine if tCP should be considered an additional controlling parameter (in addition
to j, q, and tM ) and to determine if small increases in tCP (such that total treatment
183
time is constrained to 6-7 hours) are capable of increasing arsenic removal capacity in a
significant way.
Batch tests at j = 1.1 mA/cm2 were performed in both SBGW-1 (containing low
phosphate, average bicarbonate, and no silicate) and SBGW-2 (containing average levels
of phosphate, bicarbonate, and silicate for Bangladesh groundwater). The additional
phosphate and silicate in SBGW-2 lead to a 54% decrease in arsenic removal capacity
relative to the WHO limit. This is consistent with previous studies of by Roberts et al.
(2004) of competitive adsorption between arsenic and phosphate or arsenic and silicate
during coprecipitation with Fe[III] salts.
No single set of operating parameters was optimal in terms of both arsenic
removal capacity and treatment time. Based on batch tests, two sets of parameters with
different advantages were chosen. The parameter set with the highest overall arsenic
removal capacity(i.e. the lowest operating costs) was j = 0.02 mA/cm2, q = 25 C/L,
tCP = 416 min (∼ 7 hrs), tM = 60 min. Total treatment time (8 hrs) was just above the
target for a community water system (6-7 hrs). Constraining the treatment time to be
around 2 hours, the highest arsenic removal capacity was achieved with j = 1.1 mA/cm2,
q=150 C/L, tCP = 69 min, tM = 60 min. Both of these parameter sets have advantages
(the first in terms of operating costs, and the second in terms of treatment time) and
choosing between them would require more detailed knowledge on the importance of
reduced operating costs compared to treatment time.
In order to explore the significance of producing iron (hydr)oxides in-situ with
arsenic during ECAR and study adsorbent aging, experiments were performed using
184
iron (hydr)oxides that were not generated in-situ. Post-synthesis ECAR-generated iron
(hydr)oxides adsorbent, or PS-EGA adsorbent, was created by running the ECAR dosing
process in synthetic Bangladesh groundwater containing no arsenic. Freshly made PS-
EGA was able to reduce initial concentrations of ∼ 600 µg/L (including As[III] and
As[V] in a 1:1 ratio) to 37 µg/L after 60 minutes of contact and 22 µg/L after 48 hours
of contact. ECAR treatment under comparable conditions was able to reduce arsenic to
10 µg/L after 60 minutes of mixing. The low arsenic concentration implies that most
of the initial 300 µg/L of As[III] was removed with PS-EGA without electrochemical
oxidation. This As[III] could be forming complexes with PS-EGA directly, or being
oxidized in parallel with Fe[II] oxidation by dissolved O2. Significantly however, the
adsorbent was not able to reduce total arsenic levels to below the WHO limit (10 µg/L),
indicating that the in-situ generation of EGA during ECAR is significant to its arsenic
removal capability.
Adsorption onto freshly prepared PS-EGA was compared to adsorption onto
PS-EGA allowed to age for 60 minutes after being generated (aged-PS-EGA). Removal
capacity (as µg-Asrem/mg-Fe 3+) was 33% lower for aged-PS-EGA than fresh-PS-EGA
after 60 minutes of contact and 13% lower after 48 hours (Asfinal = 100 µg/L). This
indicates that removal capacity of EGA is highly sensitive to aging. This is likely due
to the formation of aggregates. Once aggregates form, arsenic oxyanions must diffuse
into them to find available sites. It is possible that the adsorption capacity of aged-
PS-EGA could catch up with that of fresh-PS-EGA given enough time for diffusion
into aggregates to occur. However, given the constraints on treatment time proposed in
185
Chapter 1, attempting to form PS-EGA at a central location and shipping the adsorbent
to satellite treatment sites in place of on-site ECAR treatment is not advised.
Electrochemical oxidation14 of As[III] to As[V] and subsequent adsorption of
As[V] onto EGA has been proposed as the mechanism of effective As[III] removal in elec-
trocoagulation. Measurements of As[III] concentration during ECAR treatment indicate
that the arsenic removal capacity of ECAR for As[III] is approximately 3 times lower
than the removal capacity for As[V]. This is consistent both with (1) EGA having a lower
affinity for As[III] (thereby requiring a higher dosage) or (2) slow oxidation of As[III],
requiring more current processing time rather than more charge. Freshly made PS-EGA
(see previous paragraphs) was able to remove 90% of added As[III] after 60 minutes of
contact in SBGW containing competitive ions such as phosphate, silicate, bicarbonate,
and As[V]. No electrodes were present or external voltage applied during the contact
time, preventing electrochemical oxidation of As[III]. This implies that either (a) EGA
has a significant affinity for As[III] on the timescale of ECAR treatment, even in the
presence of competing ions, or (b) As[III] is being oxidized by something other than the
anode, perhaps in parallel with Fe[II] oxidation due to dissolved O2. Final As[III] con-
centrations are still lower after ECAR treatment than contact with PS-EGA, implying
that either electrochemical oxidation is significant in reducing low arsenic concentrations
down to less than 10 µg/L, or perhaps that EGA generated in-situ with As[III] has a
higher arsenic removal capacity than PS-EGA.
Transient voltammetry was performed on the iron electrode at ν = 0.1 mV/s in14Technically, all redox reactions are electrochemical, however, here I am referring to electrochemical
oxidation caused by the external voltage applied.
186
0.1 M KClO4 (plain perchlorate) solution and 0.1 M KClO4 plus synthetic Bangladesh
groundwater (SBGW-1 recipe). A comparison of the resulting oxidation scans at low
potentials revealed an anodic shift of approximately 50 mV in SBGW-1 compared to
plain perchlorate solution, resulting in a large loss of current. This suggests that species
in the SBGW interfere with the dissolution processes on the iron electrode. This resulting
increase in the overpotential required to reach a given current density in SBGW will add
to the energy consumption of ECAR treatment. Since the overpotential is less severe at
higher current densities, this could give high current densities some advantage over low
current densities in terms of energy consumption. This could help offset disadvantages
due to higher operating current.
Electrode passivation, a common problem during electrocoagulation with alu-
minum electrodes, is not seen within the range of the scan (up to current density j =
10 mA/cm2). The lack of passivation in iron over the current density region that is most
effective for arsenic removal is promising in terms of electrode longevity, maintenance
requirements, and the ability to operate in monopolar mode.
The polarization scan revealed that the electrode potential over the entire range
of scanned current density (i.e. j ≤ 10 mA/cm2) is well below the thermodynamic
equilibrium potential for oxygen evolution. This rules out oxygen evolution as a cause
of the decrease in arsenic removal capacity seen at higher current density. The onset of
oxygen evolution also limits the highest current density that could potentially be effective
in ECAR. Because that limit has not yet been reached, higher current densities should be
explored to look for possible increases in arsenic removal capacity (if the composition of
187
EGA changes favorably) that may occur before reaching the oxygen evolution potential.
Since higher current densities can deliver more charge in less time, this could potentially
provide a very favorable parameter space for arsenic removal.
A comparison of the polarization scan in SBGW to thermodynamic equilib-
rium potentials for relevant iron species revealed that the lowest current density (j =
0.02 mA/cm2) is likely to produce iron (hydr)oxides based on Fe[II], while the higher
current densities are likely to produce iron (hydr)oxides based on Fe[III]. The high re-
moval capacity seen at low current density suggests that Fe[II] species may have a higher
affinity for arsenic than Fe[III] species. A comparison of the polarization scan in SBGW
to thermodynamic equilibrium potentials for relevant arsenic species revealed that As[V]
reduction to As[III] may be thermodynamically favorable at the lower current densities
tested. The fact that arsenic removal capacity is higher at low current density suggests
that electrochemical oxidation of As[III] may not be essential for effective and efficient
As[III] removal in ECAR. This is consistent with the relatively high adsorption of As[III]
achieved by post-synthesis EGA after 60 minutes of contact in SBGW-2.
The arsenic removal capability of ECAR using 0.1µm filtration was compared
to that of ECAR using sedimentation (i.e. quiescent settling followed by decantation).
Dosed and mixed synthetic Bangladesh groundwater samples (SBGW-1) were left to set-
tle over a period 2 - 6 days and compared to samples filtered immediately after mixing.
For a settling time of less than 3.1 days, sedimentation samples contained 10 - 64 µg/L
more arsenic than 0.1 µm filtered samples. After 3.8 days of settling, sedimentation
samples contained approximately 20 µg/L less arsenic than 0.1 µm filtered samples. The
188
increase in arsenic removed is probably due to the additional complexation time allowed
by the settling period. Repeated sampling after 6 days of settling showed no additional
decrease. Attempts to increase the settling rate using magnets were unsuccessful, sug-
gesting that the EGA sludge is not very magnetic. Sedimentation tests using SBGW-2
water produced similar results. The difference in arsenic concentration between sedimen-
tation samples and 0.1 µm filtered samples approached zero after 1.9 days of settling.
Increasing the mixing time by 30 minutes before quiescent settling accelerated the initial
stages of settling, but 1.9 days was still required before sedimentation results began to
approach 0.1 µm filtered results. These tests imply that sedimentation cannot replace
0.1 µm filtration without requiring water storage for up to 2 days or the addition of a
coagulant.
189
Chapter 4
Characterization of Reaction
Products
4.1 Introduction
Electrochemical arsenic remediation (ECAR) produces arsenic-laden iron (hydr)oxide
sludge (EGA) during the treatment process. EGA sludge is composed of the reaction
products formed during electrolysis (also called coulombic dosing) and post-electrolysis
mixing1. Characterization of these reaction products is important to (1) characterize the
particles that must be separated from potable water in the filtration step, (2) charac-
terize the waste byproduct associated with ECAR treatment, (3) explore the reactions
and arsenic removal mechanism in ECAR, and (4) explore the effect of ECAR operating
parameters on arsenic removal. Each will be described in turn.1For a review of the three stages in ECAR treatment, see Section 3.2.5.
190
ECAR treatment involves a coulombic dosing stage, in which EGA particles
are generated in arsenic-contaminated solution, a mixing stage, in which complexation
between arsenic and EGA along with coagulation occurs, and a separation or filtra-
tion stage, in which arsenic-laden EGA is separated from potable water. In laboratory
scale batch tests, filtration was achieved using 0.1µm pore size membrane filters (see
Chapter 3). However, in rural Bangladesh, a low-cost alternative is desirable. Previous
experiments to separate particles through simple settling and decantation (Chapter 3)
revealed that > 3 days of settling is required to achieve the same level of particle separa-
tion as 0.1µm filtration. A more detailed study of alternative low-cost options is needed,
for which it is useful to characterize the average EGA particle size, shape, and homo-
geneity. Scanning electron microscopy (SEM) is a technique capable of imaging surface
topography with high resolution (∼5 nm). SEM images can be used to estimate the
primary aggregate size, particle shape, and homogeneity of an EGA sample. A simple
comparison of filtrate arsenic concentration after membrane filtration with different pore
sizes can help constrain the cluster size of EGA as well.
Arsenic-laden EGA sludge must be disposed of as a waste product after ECAR
treatment. The amount and chemical composition of the waste produced are key to un-
derstanding safe disposal mechanisms and potential disposal costs in Bangladesh. Chem-
ical analysis of dissolved EGA sludge can provide information about the arsenic content
as well as the As/Fe ratio of the waste. The total sludge produced after ECAR treatment
of synthetic Bangladesh groundwater can be weighed to estimate the waste per coulomb
expected in Bangladesh. The strength and stability of the surface complexes between
191
arsenic and EGA are also relevant to safe waste disposal. X-ray absorption fine structure
(EXAFS) spectroscopy has been widely used to study the bonding mechanism of arsenic
by geosorbents, such as iron (hydr)oxides (Fendorf et al., 1997; Paktunc et al., 2003).
Inner sphere complexation between As and Fe hydroxides has been shown by EXAFS
(Waychunas et al., 1993; Manceau, 1995; O’Reilly et al., 2001). EXAFS also showed
that bidentate binuclear complexation was the major bonding mechanism for arsenate
adsorption onto goethite (Fendorf et al., 1997; Waychunas et al., 1993; Manceau, 1995;
O’Reilly et al., 2001). If a similar type of strong bonding exists between arsenic and
EGA, this could provide initial evidence that As would remain immobilized in EGA
waste.
An understanding of the arsenic bonding structure to EGA is also relevant to
understanding the dominant arsenic removal mechanism in ECAR. EXAFS spectroscopy
can distinguish between surface complexes of arsenic onto EGA and the formation of a
solid phase, such as FeAsO4 (scorodite) (Foster, 2003). The primary location of arsenic
removal (i.e. on the surface layer of the electrode versus within the bulk solution)
could also be determined using EXAFS spectroscopy. Iron EXAFS spectroscopy and
X-ray absorption near edge spectroscopy (XANES) could also be used to characterize
the dominant iron (hydr)oxides structure of EGA. Fe atoms are 6-fold coordinated in
iron oxyhydroxide (Manceau & Combes, 1988). The 3D chains are made up of single or
multiple chains of edge sharing octahedra that in turn may be joined along their lengths
by sharing faces (e.g. Hematite), edges (e.g. lepidocrocite), or corners (goethite). Fe-Fe
distances are roughly characteristic of a given polyhedral linkage (Manceau & Combes,
192
1988). Fe-XANES and EXAFS has been used to determine the oxidation state of iron
compounds and distinguish between mineral classes such as oxides and oxyhydroxides
(O’Day et al., 1994).
In Chapter 3, ECAR batch tests in synthetic Bangladesh groundwater revealed
some difference in arsenic removal capacity with current density. Polarization curves
ruled out oxygen evolution as the primary cause. It was suggested that these differences
in removal capacity could be due to differences in EGA composition at different current
densities. It was further suggested by polarization curves and thermodynamic equilib-
rium lines that EGA generated at lower current densities (with higher removal capacity)
might be based on Fe[II] species while EGA generated at higher current densities might
be based on Fe[III]-species. Fe XANES spectroscopy could determine the primary oxi-
dation state of Fe in EGA. Fe EXAFS spectroscopy could be used to compare the EGA
structure generated at different current densities and look for differences. As EXAFS
spectroscopy is sensitive to the bonding structure of the iron compounds preferred by
arsenic. As EXAFS of EGA generated at different current densities could reveal if ar-
senic prefers bonding to different components of EGA even if the primary component of
EGA is unchanging.
4.1.1 Research objectives
The primary goal of this research was to characterize the EGA reaction products
of ECAR generated at different current densities using chemical analysis, SEM, XANES,
and EXAFS spectroscopy. The purpose of this characterization was fourfold: (1) to
193
physically characterize the particles that must be separated from potable water in the
filtration step of ECAR, (2) to characterize the waste products associated with ECAR
treatment for safe disposal, (3) to determine the bonding structure of arsenic complexed
to EGA responsible for arsenic removal in ECAR, and (4) to explore a difference in EGA
composition as the primary cause of arsenic removal capacity difference seen at different
current densities.
Specific Research Objectives
1. Characterize the primary aggregate size, cluster size, particle shape, and homo-
geneity of EGA for the purpose of designing a low-cost filtration mechanism in
Bangladesh.
2. Measure the As content and Fe/As ratio of arsenic-laden waste produced in ECAR
at various current densities and use this to estimate the amount of waste per
coulomb generated in ECAR.
3. Compare the iron (hydr)oxides structure of EGA produced at different current
densities (EXAFS).
4. Compare the arsenic-EGA bonding structure in arsenic-laden EGA at different
current densities (EXAFS).
5. Compare the iron (hydr)oxides structure of EGA to known reference iron oxyhy-
droxide and iron hydroxide compounds (EXAFS).
6. Compare the arsenic-iron (hydr)oxides bonding structure in arsenic-laden EGA to
194
the known bonding structure of arsenic adsorbed to various iron (hydr)oxides and
scorodite (EXAFS).
4.2 Methods
4.2.1 Arsenic breakthrough as a function of membrane pore size
A rough limit on the cluster size of EGA particles was measured during the
filtration stage of ECAR using membrane filters with progressively larger pore sizes, from
0.1 - 0.8µm. The membranes used were Millipore polyvinylidene fluoride hydrophilic (
0.1µm), Pall Supor 450 ( 0.45µm) and Pall supor 800 (0.8 µm). In each case the pore
size is the measure of the largest pore 2. The sample used was SBGW-1 recipe synthetic
Bangladesh groundwater treated at j = 0.70 mA/cm2, q = 14.8 C/L, tCP = 3 min,
tM = 60 - 135 min. The arsenic concentration (measured via Quick Test described in
Section 3.2.1) in 0.1µm pore size filtrate was stable (within 5 µg/L) between tM = 40 min
and tM = 60 min. All tests were duplicated.
4.2.2 Scanning Electron Microscopy (SEM)
Scanning electron microscopy (SEM) was used to image the arsenic-laden EGA
sludge generated by ECAR treatment in synthetic Bangladesh groundwater. SEM im-
ages the sample surface by scanning it with a high-energy beam of electrons. The elec-
trons interact with atoms on the surface of the sample providing information about the2Pore size is determined using a bubble point test, in which liquid is held in pores by surface tension
and the minimum pressure required to force the liquid from the capillary pore structure is directly relatedto the capillary diameter.
195
topography and enabling a uniquely large depth of field in the resulting micrograph.
Arsenic-laden EGA sludge was collected on a 0.1µm membrane after ECAR treatment
at j = 1.1 mA/cm2. The sample dried in air and aged 7-9 days by the time of imaging.
A sludge coated membrane was coated in metallic gold and imaged using a Hitachi SE/N
4300 scanning electron microscope at Lawrence Berkeley National Lab (LBNL). An ac-
celerating voltage of 15kV was used to obtain images at 500x, 50,000x, and 100,000x
magnification. Energy dispersive spectroscopy (EDS) was performed on several areas
of the sample using a NORAM System Six (NSS; Thermo Scientific). Work with the
scanning electron microscope was made possible by the excellent training and direct
assistance of Dr. Eduardo Saiz of LBNL.
4.2.3 Chemical analysis of EGA sludge
During the Matrix 2 run of batch tests (described fully in Section 3.2.7), arsenic-
laden EGA sludge was collected onto 0.1µm membranes and dried3. The dry powder was
scraped into glass test tubes and stored in open air. In one case (j = 0.020 mA/cm2),
powder was collected from the iron electrode surface (after drying). Powder was lightly
tapped off of the electrode rather than scraped in order to avoid scraping metallic iron
flakes from the electrode itself.
All powder samples were dissolved in a 1.5:1 solution of 38%-HCl acid: deion-
ized water. The powder was fully dissolved after approximately 4 hours, and a 1:50
dilution was sent for ICPMS analysis of aqueous iron and arsenic (ICPMS described in3Drying occurred in a glove box purged with N2 gas in conjunction with preparation of XAS samples
(described in Section 4.2.4. However the powder was later exposed to air.
196
Section 3.2.1). Reproducibility of this procedure was tested by sending four samples
prepared from the same powder, in each case dissolving a different amount from 25 -
73 mg per 10 mL solution. The resulting Fe/As mass ratios were within 4% of eachother.
In one case, the total mass of sludge produced was measured using a precision
balance (up to 0.0001 g)4. However, ∼ 1 mg may have been lost due to flaking limiting
the precision of the measurement.
4.2.4 X-ray Absorption Spectroscopy (XAS)
4.2.4.1 Overview
X-ray Absorption Spectroscopy (XAS) consists of recording the absorption of
X-rays by target atoms as a function of the X-ray beam energy (eV) to create an XAS
spectra. Spectral features near the absorption edge appear within the X-ray Absorp-
tion Near-Edge Structure (XANES) spectrum. Farther from the absorption edge, an
oscillatory signal is produced by the backscattering of ejected photoelectrons by atoms
immediately surrounding the target atom. The oscillatory signal is called the Extended
X-ray Absorption Fine Structure (EXAFS) spectrum. Given the short mean free path
of the ejected photoelectron (a few angstroms for a kinetic energy of a few hundred eV)
structural information can be discerned form the first atomic shells surrounding the X-
ray absorber. This ability makes EXAFS analysis a unique tool for elucidating the local
structure of poorly crystallized particles and surface complexes (Charlet & Manceau,
1993). The EXAFS spectrum, χ is a function of the photoelectron wave vector, k, where4The total mass of sludge was not weighed in every case because some sludge was removed to prepare
XAS samples before drying - described in Section 4.2.4 below.
197
k is related to the kinetic energy of the photoelectron (E) by:
k = [2mh2 (E − E0)]1/2 (4.1)
where E0 is the threshold energy of the photoelectron at k = 0, and m is the mass of
the electron. A typical form of the EXAFS function for a single set of NR atoms at a
distance R from the absorbing atom is given by (O’Day et al., 1994):
χ(k) =∑r
NRS20
|feff (π, k,R)|kR2
sin(2kR+ 2δc + Φ)e−2σ2k2e−2R/λ (4.2)
where feff (π, k,R) is the effective curve-wave backscattering amplitude of the scatterer
(Teo & Joy, 1981), δc and Φ are phase shifts for the absorber and backscatterer respec-
tively, S20 is a many-body amplitude reduction factor, σ2 is a Debye-Waller term (or a
mean-square relative displacement) assuming a harmonic oscillator, and λ is the mean
free path of the electron. Theoretical functions for the phase (δc and Φ) and amplitude
( feff (π, k,R)) are generated from ab initio theory by the program FEFF (Rehr et al.,
1991) using model compounds of similar structure. This allows extraction of the values
NR, R, and σ2 from the EXAFS function.
For this work, EXAFS and XANES spectra were collected over four short runs
at two synchrotron facilities. The sample preparation procedure was refined and im-
proved with each run, as described below.
198
4.2.4.2 Iron XANES and XRF-maps
Sample preparation
Fe XANES samples were prepared from the EGA sludge and the surface of the
iron electrode after ECAR treatment5 of SBGW-1 synthetic groundwater6 at various
operating conditions (listed in Table 4.1). The initial pH of the synthetic groundwater
before ECAR treatment was pH = 7.80, slightly higher than the average pH of SBGW-1
used in Matrix 1 batch tests7. Sludge samples were collected after the ECAR mixing
stage on polyvinylidene fluoride membrane filters (Millipore; 0.1µm pore size) and dried
in air. Electrode surface layer rubbing samples were collected by pressing Kapton tape to
the wet electrode surface immediately after ECAR dosing. All EGA sludge were allowed
to dry in air (individual sample ages are listed in Table 4.1), though it should be noted
that in several cases, samples were prepared at the beamline and placed in the beam wet
and within 5 minutes of completing ECAR treatment (the listed age for these samples
is 0 days).
Note that ECAR treatment at j = 5.0 and 0.02 mA/cm2 was performed in a
much smaller electrochemical cell than the than the 850 mL-separated cathode used for
Matrix 1 experiments. The anode was formed from an iron wire (diameter 0.18 cm)
coated in epoxy resin and sanded at one end - exposing an iron disk of area 0.025 cm2.
The cathode was a platinum wire coil approximately 2 mm away from the anode. The cell
was a small glass test tube containing 220 µL of solution, forcing the same A/V ratio of5The ECAR treatment procedure is detailed in Section 3.2.5. The ECAR operating parameters are
summarized in Section 1.4.1.6The composition of SBGW-1 is given in Table 3.1.7Matrix 1 batch tests are presented in Chapter 3.
199
the 850 mL-separated cathode cell used in Matrix 1 experiments (A/V = 0.118 cm−1).
The solution was not continuously stirred during the treatment or the mixing stage,
though the test tube was lightly shaken periodically during both processes. Before ECAR
treatment, the anode was conditioned for 6 minutes at a constant current of -1 nA to
electrochemically reduce any oxide layer that may have formed on the electrode while
sitting in air. The small cell was designed to facilitate grazing incidence spectroscopy,
described below.
Data collection
Fe k-edge XANES spectra and spatially-resolved X-ray fluorescence (XRF)
maps were collected under ambient conditions on beamline 10.3.2 at the Advanced Light
Source (ALS; Berkeley, CA) with a 7-element germanium detector in fluorescence mode
(Marcus et al., 2004). Ring conditions were 200-400mA and 1.9 GeV. Monochromator
energy calibration was set with an Fe(0) foil at 7110.75 eV. XRF elemental maps were
collected at an incident energy of 12067 eV (i.e. 200 eV above the As K-edge) with a pixel
size of 5 x 5 µm and a dwell time of 100 ms per pixel. Maps were deadtime corrected and
registered using custom beamline software developed by M. Marcus (ALS). The maps
were used to locate regions of interest for collection of Fe XANES spectra. Experimental
spectra were deadtime corrected and energy calibrated using custom beamline software
developed by M. Marcus (ALS).
200
A note on grazing incidence spectroscopy
The original purpose for choosing the 10.3.2 micro-beam at ALS was to perform
grazing incidence spectroscopy on the electrode surface immediately after ECAR dosing.
Grazing incidence spectroscopy takes advantage of the fact that the penetration depth
of the X-ray beam is controlled by the glancing angle - if the angle can be controlled
enough to cause a total reflection, the penetration depth can be suppressed to the order
of a few nm (Sato et al., 2005). The beam footprint is lengthened by the grazing angle,
making a small beam such as the micro-beam 10.3.2 at ALS, ideal to obtain maximum
flux. Performing spectroscopy on the electrode itself immediately after treatment was
intended to minimize disruption and aging of the surface layer, leading to a more accurate
picture of the surface layer during the ECAR dosing stage.
The attempt to take spectra at grazing incidence ultimately failed as the signal
from the surface layer was too weak to compete with the Fe(0) signal of the electrode
itself. A steeper grazing angle (leading to a smaller penetration depth) was not able
to compensate because the electrode surface was neither uniform enough nor smooth
enough for the entire beam area to penetrate to the same depth at a steep angle. One
suggestion for looking at the surface directly is to sputter a uniform and thin layer of
iron onto an inert substrate and perform in-situ ECAR treatment while measuring the
EXAFS spectra. Because X-rays penetrate an appreciable distance through water, it
is possible to perform surface studies at the electrolyte-solid interface (Tang & Furtak,
1991). Building and testing such a cell was unfortunately beyond the scope of this work.
This section is included for the benefit of future investigators.
201
Table 4.1: ECAR operating conditions used to generate XAS samples. ECAR operatingconditions (summarized in Section 1.4.1) include current density, charge density, currentprocessing time (tCP ), and mixing time (tM ). XAS sample preparation procedures aredescribed in Sections 4.2.4.2 - 4.2.4.5.
Spectra Current Sample Beam Charge SBGW SampleType Density Typeb Locationc Density tCP tM Reciped Agee
(mA/cm2) (C/L) (min) (min) (days)
Fe XANES 1.1 EGA sludge ALS 100 12.8 60 1 1Fe XANES 1.1 Elec SL ALS 100 12.8 60 1 1Fe XANES 5 EGA sludge ALS 100 2.83 40 1 0Fe XANES 5 Elec SL ALS 100 2.83 40 1 0Fe XANES 0.02 EGA sludge ALS 100 708 40 1 0Fe XANES 0.02 Elec SL ALS 100 708 40 1 0Fe XANES 0.07 EGA sludge ALS 100 203 60 1 2Null As EXAFS 1.1 Elec SL SSRL 150 66.58 60 2 5Null As EXAFS 5 Elec SL SSRL 150 14.65 60 2 7As EXAFS 1.1 EGA sludge SSRL 175 74.98 60 2 4As EXAFS 0.02 EGA sludge SSRL 75 1165.25 60 2 5As EXAFS 100 EGA sludge SSRL 175 9.2 60 2 7As EXAFS 5 EGA sludge SSRL 175 16.33 60 2 3Fe EXAFS 1.1 EGA sludge ALS 175 25.58 60 2 2Fe EXAFS 0.02 EGA sludge ALS 75 452.92 60 2 4Fe EXAFS 100 EGA sludge ALS 175 3.17 60 2 4Fe EXAFS 5 EGA sludge ALS 175 5.63 60 2 2
a Null As EXAFS refers to attempts to generate As k-edge EXAFS spectra that failed to produce any As signal.b Sample type can be either EGA Sludge, referring to samples collected from the bulk electrolyte, or Elec SL,
referring to samples collected from the electrode surface layer after ECAR treatment.b Beam location is either ALS (Advanced Light Source - Berkeley, CA) or SSRL (Stanford Synchrotron Radi-
ation Laboratory- San Jose, CA).d Synthetic Bangladesh Groundwater (SBGW) recipes 1 and 2 are described in Table 3.1.e Sample age refers to the time between sample preparation and spectra collection.
202
4.2.4.3 Null arsenic EXAFS attempts
Null arsenic EXAFS attempts refer to several attempts to generate As k-edge
EXAFS spectra from electrode surface layer scrapings that failed to produce any As
signal. These results are significant in constraining the location of arsenic removal in
ECAR (i.e. on the electrode surface or within the bulk electrolyte), and as such, the
sample preparation procedure and attempted data collection are described here.
Sample preparation
Electrode surface layer samples were collected after ECAR treatment of SBGW-
2 synthetic groundwater8 during Matrix 2 batch experiments9 at various operating con-
ditions (listed in Table 4.1). After the dosing stage of ECAR treatment, the iron anode
was immediately transferred to an anaerobic chamber for drying. Within the anaerobic
chamber, the dry electrode surface layer was carefully scraped with a clean Xacto knife
blade down to the metal electrode10. In the case of dosing at j = 0.02 mA/cm2, the
electrode surface was caked with a thick (∼0.5 mm) layer of orange powder that easily
flaked off after drying without the use of a knife blade. Electrode surface layer scrapings
were packed into glass capillaries with a 2 mm outer diameter. The estimated thick-
ness of the glass itself is 0.25 mm (by eye11). Capillaries were capped with epoxy resin
and stored at ambient temperature in the anaerobic chamber until transfer to the beam8The composition of SBGW-2 is given in Table 3.1.9Matrix 2 batch tests are presented in Chapter 3.
10Note that in most cases, this procedure is not appropriate for Fe k-edge spectra because flakes ofFe(0) from the electrode or the stainless steel knife blade can easily contaminate the sample.
11The exact thickness is unknown because the capillaries were hand-pulled rather than purchased froma manufacturer.
203
between 5-9 days after collection (individual sample ages are listed in Table 4.1).
It should be noted that after a reconnaissance trip to the beam, capillaries
were found to be to too long for the sample chamber and the samples were repacked into
new shorter glass capillaries (same outer diameter) 2 days before data collection. This
relevant because during the repacking, it was noted that the electrode surface scrapings
were far more magnetic than they had been during the first packing. This is a sign that
the iron (hydr)oxides were aging into more crystalline magnetite or maghemite (both
magnetic) during anaerobic storage.
Data collection attempts
Arsenic k-edge EXAFS collection was attempted with an unfocused beam (beam-
size 2 x 20 mm) under ambient conditions at beamline 11-2 at the Stanford Synchrotron
Radiation Laboratory (SSRL; San Jose, CA). Ring conditions were 50 - 100 mA at 3GeV.
The beam was detuned 30% to avoid harmonics. Energy calibration was performed with
an arsenic metal foil at 11867 eV. Beam energy was selected using a Si(220) monochro-
mator and fluorescence signals were generated using a 30-element Germanium detector.
In all cases, transmission and fluorescence signals were measured simultaneously. Stoller
slits and Ge filters were used in front of the fluorescence detector to minimize the role of
Compton scattered primary X-rays. In each attempt to find an arsenic signal, the beam
was physically centered on the sample using the easily detectable Fe peak and the slits
were opened as wide as possible (2 x 20 mm) at a beam energy 200 eV above the As
k-edge. Various combinations of filters (i.e. Al and Ge) were used to help resolve any
204
possible As signal. Successful As k-edge EXAFS spectra were measured for the electrode
scrapings from j = 0.02 mA/cm2, proving that the setup was in proper working order
for As detection12.
4.2.4.4 Arsenic k-edge EXAFS
Sample preparation
EGAsludge samples were collected after ECAR treatment of SBGW-2 synthetic
groundwater13 during Matrix 2 batch experiments at various operating conditions (listed
in Table 4.1). After the ECAR mixing stage, sludge from ∼ 1 liter of dosed and mixed
solution was collected on a filter membrane ( 0.1µm pore size; membrane filtration took ≈
1 hour). Once filtration was complete, the wet sludge was transferred into the anaerobic
chamber while submerged in filtrate solution (keeping the sludge wet during the 10
minute transfer into the chamber). Once in the chamber, wet sludge was scraped from
the membrane onto Kapton film and formed into a pill shape in order to concentrate
the arsenic as much as possible. A smaller piece of Kapton film was placed on top and
the edges sealed with epoxy resin to maintain a wet paste consistency. Two identical
paste preparations were prepared for each sludge sample. After 5 minutes (to allow the
epoxy to dry), the sealed paste samples were sealed inside 20 mL glass jars (to slow
oxygen contamination through the Kapton film) and transferred from the anaerobic
chamber into a 1C storage unit. The 20 mL glass jars were later placed inside 150 mL12As EXAFS spectra for j = 0.02 mA/cm2 electrode scrapings were later remeasured using a different
sample preparation procedure, described in Section 4.2.4.4.13The composition of SBGW-2 is given in Table 3.1.
205
plastic bottles (within the anaerobic chamber) to act as an additional barrier to oxygen
contamination during storage. Samples were stored for 3-7 days (individual sample ages
are listed in Table 4.1).
Data collection
Arsenic K-edge EXAFS spectra were taken on beamline 11-2 at SSRL under the
same beam conditions described in Section 4.2.4.3. However, in this case, samples were
removed from protective jars and placed immediately into a small anaerobic cell where
they remained during the EXAFS scan. All wet paste samples appeared wet (by eye and
touch) before and after EXAFS spectra collection. Spectra were collected using a beam
size of 2 x 20 mm to a k-value of 14 A−1. Initially, beam harmonics were reduced using
grazing incidence cut-off reflection mirrors to eliminate higher order radiation. However,
it was discovered during the run that spectral data above k= 11 A−1 was extremely
noisy and non-reproducible between scans. The mirrors were removed and the beam was
detuned 30% to reduce harmonics, as was the case during the first SSRL run (described
in Section 4.2.4.3). This reduced the noise considerably and led to reproducible data.
All data presented here used detuning, however this note was included to inform future
EXAFS data collectors.
206
4.2.4.5 Iron k-edge EXAFS
Sample preparation
Sample preparation for Fe k-edge EXAFS was the same as for As k-edge EX-
AFS (Section 4.2.4.4), with the following exceptions: (1) Saran wrap was sandwiched
inside the Kapton film surrounding each sample to further protect against oxygen con-
tamination, (2) sealed Kapton/Saran film samples were placed inside 20 mL glass jars
and 150 mL plastic jars immediately after being formed inside the anaerobic chamber -
neither jar was opened until just before EXAFS data collection, and (3) instead of scrap-
ing the wet sludge off of the membrane, a small piece of wet sludge-covered membrane
was cut into strips and sealed within the Kapton/Saran wrap film. The last modification
was done to provide a more uniform thin sludge layer14. All samples were stored at 1C
until EXAFS data collection.
Data collection
Iron k-edge EXFAS spectra were measured at beamline 10.3.2 at ALS. Ring
conditions were 200 - 400 mA at 1.9 GeV. The beam was detuned 20% to avoid harmonics.
Energy was calibrated to Fe(0) foil at 7110.75 eV. Spectra were collected using a beam
size of 100 x 10 µm to 100 x 20 µm. Individual scans collected at the same location were
examined for changes in line shape and peak position and no beam radiation damage
was observed. Spectra were collected to a k-value of 14 A−1. However, it was discovered
during EXAFS collection that data past k ∼11 A−1 had very low signal-to-noise and was14Note also that previous samples were concentrated to get a better signal from trace arsenic. Iron
was much more abundant in the samples, and therefore a thin layer provided plenty of iron signal.
207
not reproducible between scans. This may have been due to the wet samples settling in
gravity during the scan, effectively changing the beam penetration depth across the scan.
The EXAFS from each scan were compared by eye and the average data was cut off to
an appropriate k-value such that data was consistent across scans. Acceptable k-values
were between k = 10.5 and 11.2 A−1 Experimental spectra were deadtime corrected and
energy calibrated using custom beamline software developed by M. Marcus (ALS).
It should be noted that the micro-beam available at 10.3.2 was not ideal for
these samples. The EGA sludge is quite uniform, negating the spacial resolution power
of a micro-beam. In addition, the increased flux produced by a larger beam would allow
spectra to be taken much more quickly and possibly without some of the problems at
high k discussed above. Beamline 10.3.2 was used because time on a bulk beam was not
otherwise available15.
4.2.4.6 Reference spectra
Reference Fe XANES spectra for synthetic iron (hydr)oxides, including 2-line
ferrihydrite, 6-line ferrihydrite, goethite, lepidocrocite, akaganeite, magnetite, and maghemite,
along with Fe metal foil, were collected at beamline 10.3.2 at ALS by Dr. Matthew Mar-
cus (ALS) and provided via personal communication. Spectra were received averaged,
deadtime corrected, and energy calibrated.
Reference Fe EXAFS and Fe XANES spectra for synthetic samples of scorodite
(FeAsO4·2 H2O), carbonate green rust ((Fe[II]4Fe[III]2)(OH)12CO3), and sulfate green
15Dr. Matthew A. Marcus and Ms. Sirine Fakra deserve thanks for making the micro-beam work aswell as it could under non-optimal conditions.
208
rust ((Fe[II]4Fe[III]2)(OH)12SO4) were provided by Dr. Peggy O’Day (Arizona State
University). Sample preparation and data collection are described in O’Day et al. (1994).
Spectra were received averaged, deadtime corrected, and energy calibrated.
Reference As EXAFS spectra for arsenate (As[V]) adsorbed onto 2-line ferrihy-
drite, arsenate adsorbed onto goethite, and scorodite were provided by Dr. Andrea Foster
(US Geological Survey). Sample preparation and data collection are described in Foster
(1999). Spectra were received averaged, deadtime corrected, and energy calibrated.
While all of the above authors were kind enough to offer me their spectra for my
thesis work, they have neither read nor approved of the analysis based on comparisons to
the above spectra, and thus it should not be taken to reflect their professional opinion.
4.2.4.7 XAS data analysis
Spectra averaging was performed using in SIXPack software (Webb (2003);
SIXPack v2.3.1) for arsenic EXFAS spectra and using ”EXAFS editor.exe” developed by
M. Marcus (ALS) for iron XANES and iron EXAFS spectra. In the case of iron EXAFS
spectra, each scan in the average was viewed side by side to determine an appropriate
data cut off point based on reproducibility of the signal between scans (see discussion in
Section 4.2.4.5). All averaged, unsmoothed EXAFS spectra were background-subtracted
by a linear fit through the pre-edge region and a cubic spline through the spectrum above
the absorption edge using SIXPack. A threshold energy (i.e. the energy at which k, the
photoelectron wave vector, equals zero) of 7132 eV was used for iron and 11873 eV was
used for arsenic. XANES spectra were normalized using the height of the edge step just
209
above the absorption maximum using ”EXAFS editor.exe.” All reference spectra from
outside sources were re-processed using the same methods as sample spectra used in
comparisons.
4.3 Results and Discussion
4.3.1 Limit on the average EGA cluster size
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.6
0.7
0.8
0.9
1.0
Membrane Pore Size (microns)
Nor
mal
ized
Filt
rate
Con
cent
ratio
n
Figure 4.1: Arsenic concentration measured in filtrate (normalized to the highest averagevalue) after ECAR-treated sample water passed through membranes of different poresize. Errors represent the variability between duplicate tests.
A rough lower limit on the average size of EGA clusters can be determined
through filtration tests by progressively increasing the pore size of the filter membrane.
Filtrate arsenic concentrations after filtration with 0.1 - 0.8µm pore size membranes
were measured in SBGW-1 after ECAR treatment at j = 0.70 mA/cm2, q = 14.8 C/L,
210
tCP = 3 min, tM = 60 - 135 min (Figure 4.1). A small decrease in the effluent arsenic
was observed, though this decrease was smaller than the variability between duplicate
tests (indicated by error bars), resulting in no significant difference across pore sizes. A
decrease in effluent arsenic is also unlikely, since a larger pore size should never lead to
the passage of fewer particles. Some room for variability was introduced by allowing the
mixing time to vary between filtrations, as clusters are likely to get larger with increasing
mixing time. However, filtration occurred in order of decreasing pore size, and thus any
bias would tend to lead inflate effluent concentrations at higher pore size. Thus this
data, while not ideal, implies that the cluster size of EGA is greater than 0.8µm. This
result is significant largely because similar experiments reported in the literature often
use 0.45µm pore size filtration, which can now be robustly compared to batch test results
using 0.1µm filtration. It also increases the measured minimum cluster size by a factor
8, which is relevant for selecting low cost alternatives to 0.1µm membrane filtration for
a prototype device.
4.3.2 Scanning Electron Microscopy (SEM)
ECAR generates an iron (hydr)oxide sludge used to capture arsenic. The
sludge, collected on a 0.1µm polyvinylidene fluoride membrane, was imaged at 500x
magnification using SEM (shown in Figure 4.2A). At 500x magnification, the sludge ap-
pears to be a relatively uniform cake (showing cracks due to drying) dotted with small
square crystals. Marked areas were imaged at higher magnification, including one area
with a square crystal (shown at 8,000x magnification in Figure 4.2B) and one area con-
211
taining caked powder (shown at 50,000x and 100,000x magnification in Figure 4.2C and
D respectively). Energy dispersive spectra (EDS) were taken of the magnified images16,
shown in Figure 4.3B and C corresponding to Figure 4.2B and C respectively, along with
one spectra of the bare polyvinylidene fluoride membrane (Figure 4.3A; bare membrane
image not shown). The bare membrane spectra shows the characteristic Au lines cor-
responding to the gold coating used in the SEM process, and no appreciable calcium,
arsenic, or iron. The spectra of the square crystal (Figure 4.3B) shows peaks only due
to the gold coating and calcium implying that the crystals are made of calcium, presum-
ably CaCO3 (s). Notably, no iron is associated with these crystals, indicating that the
crystals are not a product of ECAR treatment. As expected, the caked powder contains
appreciable amounts of iron along with some arsenic (shown in Figure 4.3C), identifying
the magnified caked powder in Figure 4.2C and D as the iron (hydr)oxides product of
ECAR treatment (i.e. as EGA). The presence of arsenic also verifies that some arsenic
is removed in the sludge as opposed to all being adsorbed to the surface layer of the
electrode.
Magnified images of EGA (Figures 4.2C and 4.2D) show that the iron (hydr)oxides
particles are rounded, consisting of tiny spheres around 25 nm in diameter forming larger
popcorm-like clumps with a primary aggregate size of 50 - 100 nm. Only one particle
shape can be identified in areas associated with iron on the micrographs, suggesting that
the iron (hydr)oxide composition of EGA is homogeneous. Filtration tests using mem-
branes with pore sizes between 100 nm and 800 nm (Section 4.2.1) showed no increase16The spectrum of the crystal shown in Figure 4.2B was taken at a higher magnification than 8,000x
and was centered on a bare section of crystal.
212
Figure 4.2: Scanning electron micrographs of ECAR generated sludge. Image A is ofair-dried sludge collected on a membrane and magnified 500x. The two marked areasin image A are imaged at 8,000x (image B) highlighting a representative square crystaland at 50,000x and 100,000x (image C and D respectively), highlighting a representativearea of dry caked powder.
213
Co
un
ts
keV
Co
un
ts
keV
Co
un
ts
keV
Figure 4.3: Energy dispersive spectra taken during SEM imaging of (A) the barepolyvinylidene fluoride membrane (background), (B) a further magnified image of Fig-ure 4.2B (square crystal) centered on a bare section of crystal, and (C) the area shownin Figure 4.2D (caked powder). Au lines appear in all spectra due to the gold coatingsputtered onto the sample to facilitate SEM imaging.
214
in the arsenic concentration of the filtrate, suggesting that the cluster size of EGA is
greater than 800 nm.
Iron (hydr)oxides come in many different crystal shapes, and shape alone is not
a unique identifier. However, it is worthwhile to note that out of the iron (hydr)oxides
that appear at ambient temperature and pressure (e.g. goethite, lepidocrocite, aka-
ganeite, ferrihydrite, hematite, magnetite, and maghemite), only ferrihydrite and oc-
casionally magnetite and hematite appear as spheres. Goethite, known to form from
amorphous-Fe(OH)3 during aging, commonly appears in an elongated acicular shape
(Cornell & Schwertmann, 2000). Hematite, also known to form from amorphous-Fe(OH)3,
can appear as spheres, but most commonly appears as rounded or rhombohedral plate
structures (i.e. flat as opposed to spherical). Ferrihydrite is characterized by small single
spherical particles, ca. 4-6 nm in size. While not conclusive, it is interesting to note that
after 7-9 days or aging in air, the ECAR generated sludge is consistent in appearance
with ferrihydrite and does not appear to be either crystalline goethite or hematite.
4.3.3 XRF Fe-As mapping of EGA sludge
XRF mapping was done of the dry EGA sludge generated at j = 0.02, 1.1,
and 5.0 mA/cm2 at a pixel size of 2 µm. A representative map is shown in Figure 4.4
(maps at all current densities were similar). In this picture, each pixel is colored red
in proportion to the As signal and green in proportion to the Fe signal (pixels with
both arsenic and iron appear yellow). This map shows that both arsenic and iron are
uniformly spread throughout the EGA sludge. The homogeneity of arsenic and iron
215
seen here is similar to the homogeneity in particle shape and size seen in SEM images
(Figure 4.2). Taken together, these data suggest that the EGA generated for a given
current density is composed of iron (hydr)oxides of a similar structure and affinity for
arsenic.
Figure 4.4: XRF bicolor map of EGA powder. In this picture, each pixel is colored redin proportion to the As signal and green in proportion to the Fe signal (pixels with botharsenic and iron appear yellow). This map was measured for EGA generated at j =1.1 mA/cm2. Maps of EGA generated at j = 0.02 and 5.0 mA/cm2 were similar.
4.3.4 Chemical analysis of waste EGA sludge
The waste product of ECAR treatment is arsenic-laden EGA sludge (filtered out
of the bulk solution). The leftover EGA sludge after ECAR batch treatment at various
operating conditions was analyzed for Fe and As using ICPMS, listed in Table 4.2. The
ratio of As/Fe in the waste sludge depends on the coulombic dose (charge density), and
thus should only be compared between batch tests with same total charge density. Also
note that in one case (j = 0.02 mA/cm2), the electrode surface layer was collected and
tested in addition to the EGA sludge from the bulk solution.
216
Table 4.2: Chemical analysis of waste EGA sludge or electrode scrapings from ECARtreatment at various operating conditions. Sample preparation is described in Sec-tion 4.2.3.
Sample Current Total Charge Residual Total As As/Fe Fe Mass As Mass EstimatedType Density Density As Removeda Mass Ratio Fraction Fraction Wasteb
(mA/cm2) (C/L) (µg/L) (µg/L) (mg-As/ (mg-Fe/ (mg-As/ (mg/mg-Fe) mg-Waste) mg-Waste) L-treated)
EGA sludge 1.1 175 5.0 585 0.0076 0.44 0.0033 175EGA sludge 100 175 13 557 0.0079 0.54 0.0042 131
EGA sludge 30 150 14 516 0.011 0.47 0.0049 105EGA sludge 10 150 18 512 0.011 0.45 0.0050 102EGA sludge 5.0 150 11 489 0.011 0.38 0.0040 121EGA sludge 0.02 150 2.0 538c 0.014 0.39 0.0053 102
Elec Scrapings 0.02 150 2.0 538c 0.002 0.41 0.0008 ...d
a This is the total arsenic removed during ECAR treatment - i.e. the initial aqueous arsenic concentrationminus the final aqueous arsenic concentration.
b Estimated Waste uses the measured arsenic mass fraction (in mg-As/mg-Waste) and assumes all arsenicremoved for the batch was removed in either the sludge or electrode scrapings waste (i.e. it is the arsenicremoved divided by arsenic mass fraction).
c Note that this is the total arsenic removed from solution during ECAR treatment, including arsenic removedon the EGA sludge and on the electrode surface.
d The assumptions used to estimate the waste are not valid for electrode scrapings - see note b.
217
As expected, the As/Fe ratio decreases with increasing charge density due to
the additional iron dissolution. In addition, the As/Fe ratio will increase with the con-
centration of arsenic removed from the solution, creating some variation in the measured
results due to the large variation in arsenic removed. The waste characteristics for ECAR
treatment at j = 1.1 mA/cm2, q = 175 C/L provide the best upper limit for the waste
characteristics expected in Bangladesh because (1) the initial arsenic was higher than
that measured in most tube wells (> 98% of contaminated17 tube wells contain less
than 600 µg/L arsenic) and (2) the residual arsenic concentration is less than the WHO
limit of 10 µg/L, which is the treatment goal for ECAR. Solid waste for this batch test
is characterized by an arsenic concentration of 0.0033 mg-As/mg-Waste, or 3300 ppm
(1 ppm = 1 mg-As/kg-Waste), and an As/Fe mass ratio of 0.0076.
The iron mass fraction is 38% - 54% across all sludge samples (excluding the
electrode scrapings sample), consistent with an iron (hydr)oxide (for comparison, the
iron mass fraction for pure Fe(OH)3 is 52%).
The measured As/Fe mass ratio in electrode scrapings for EGA generated at
j = 0.02 mA/cm2 is 7 times less than the As/Fe ratio found in the EGA sludge. This
implies that the iron (hydr)oxide surface layer formed on the electrode has a much lower
affinity for arsenic than the EGA sludge in solution. The mass of the surface layer is
also much smaller than the mass of EGA put into solution, implying that the surface
layer affinity for arsenic would have to be very much greater than the EGA affinity for
most of the arsenic removal to occur on the electrode surface rather than in the bulk.17”Contaminated tube wells” includes all tube wells with arsenic concentrations above the WHO limit
of 10 µg/L according to the BGS (2001) - see Section 3.2.3 for more info on tube well statistics.
218
The majority of arsenic removal in ECAR is thus likely to occur in the bulk rather than
on the electrode surface.
A very rough estimate for the amount of waste expected per liter of contami-
nated water treated by ECAR (assuming a charge density of 150-175 C/L) is calculated
in Table 4.2. This estimate assumes that the arsenic mass fraction is constant through-
out the waste (a fair assumption given the homogeneity of EGA seen in SEM images and
XRF mapping) and assumes that all of the arsenic removed in the batch test occurred
in the bulk solution (i.e. on the EGA sludge). The estimated waste is 102 - 175 mg
per liter treated. In one case, the waste EGA sludge produced after ECAR treatment
of 3 liters at a charge density of 150 C/L was directly weighed, resulting in 100±1 mg
waste, or 33 mg waste per liter treated. The factor of 3 discrepancy between estimated
waste and measured waste indicates that the amount of arsenic removed from solution
is not a good predictor of waste and a better method should be explored.
4.3.5 Iron oxidation state in EGA
The peak position in XANES spectra is characteristic of the oxidation number of
the absorbing atom (Foster, 2003). In some case, features present in XANES spectra are
characteristic of a particular mineral, however, Fe[III] oxides and oxyhydroxides with Fe
in octahedral coordination are known to have few distinguishing features (O’Day et al.,
1994). The peak position can be used to distinguish between Fe[III](hydr)oxides and
Fe(II, III)(hydr)oxides, but not to identify individual minerals within these two classes
(O’Day et al., 1994). Normalized Fe XANES spectra for EGA sludge after ECAR treat-
219
ment at several different current densities (j = 0.02, 0.07, 1.1, and 5.0 mA/cm2) are
shown in Figure 4.5 along with Fe XANES spectra for various iron (hydr)oxide reference
compounds, including Fe[III](hydr)oxides: 2-line ferrihydrite (2L FH), 6-line ferrihydrite
(6L FH), goethite, lepidocrocite, and akaganeite and Fe(II, III)(hydr)oxides: green rust
(carbonate) and green rust (sulfate) (average oxidation number 2.67). The edge fea-
tures of the EGA samples are similar to the Fe[III](hydr)oxide reference compounds,
though the features are too similar to distinguish individual Fe[III](hydr)oxide minerals
(as expected). However, a line through the peak position of the EGA samples clearly
crosses the peak position of the Fe[III](hydr)oxides and not the peak position of the
Fe(II, III)(hydr)oxides (green rusts). This suggests that the oxidation of iron in EGA
generated at current densities j = 0.02, 0.07, 1.1, and 5.0 mA/cm2 is predominantly 3+
and not a mixture of 2+ and 3+.
An oxidation state of 3+ for EGA at 0.02 mA/cm2 rules out the possibility of
predominantly Fe[II]-based iron (hydr)oxides at low current density. This was suggested
by the slow scan polarization curve based on thermodynamic equilibrium potentials in
Section 3.3.10.
4.3.6 EGA iron structure - comparison to known iron (hydr)oxides
In Figure 4.6, the k3-weighted Fe EXAFS spectrum for EGA generated at j =
1.1 mA/cm2 is compared to reference spectra for iron (hydr)oxides 2-line ferrihydrite (2L
FH), goethite, and scorodite. These reference spectra were chosen as the most similar
to the sample spectrum out of a set including green rusts, hematite, magnetite, siderite,
220
7120 7140 7160 7180 7200
Energy (eV)
Akaganeite
Lepidocrocite
Goethite
6-Line Ferrihydrite
2-Line Ferrihydrite
EGA - 1.1
EGA - 0.02
EGA - 0.07
EGA - 5.0
mA/cm2
mA/cm2
mA/cm2
mA/cm2
Scorodite
Green Rust - Carbonate
Green Rust - Sulfate
No
rma
lize
d F
luo
resc
en
ce
Figure 4.5: Normalized Fe XANES spectra for EGA sludge after ECAR treatment atseveral different current densities (j = 0.02, 0.07, 1.1, and 5.0 mA/cm2) are shownalong with Fe XANES spectra for various iron (hydr)oxides reference compounds, in-cluding Fe[III](hydr)oxides 2-line ferrihydrite, 6-line ferrihydrite, goethite, lepidocrocite,and akaganeite and Fe(II, III)(hydr)oxides (average oxidation number 2.67) green rust(carbonate) and green rust (sulfate).
221
2 4 6 8 10
010
20
30
k (A)
chi(k)*
k3
k(A )-1
χ(k
)*k3
mA/cm21.12-Line Ferrihydrite
GoethitemA/cm21.1
mA/cm21.1Scorodite - FeAsO4
Figure 4.6: Iron EXAFS spectra (weighted by k3) for EGA generated at j = 1.1 mA/cm2
compared to reference spectra for iron (hydr)oxides 2-line ferrihydrite (2L FH), goethite,and scorodite. Arrows note significant differences between spectra.
222
and lepidocrocite. The four spectra in Figure 4.6 are similar in line-shape, phase, and
amplitude characteristics. As 2L FH, goethite, and scorodite all have Fe[III]-O octahedra
as the basic structural unit, one may infer that the EGA iron (hydr)oxide also has
this basic structural unit based on the spectral similarities. The spectra also possess
significant spectral differences, noted by arrows. Differences in line shape indicate that
the way in which the Fe[III]-O octahedra are linked together (Toner et al., 2008). Recall
from Section 2.3 that the basic polyhedra may be joined by faces, edges, or corners, each
with a characteristic bond length between Fe-Fe atoms affecting the shape of EXAFS.
The differences noted in Figure 4.6 indicate that these polyhedral linkages differ between
the four reference spectra and the EGA sample.
The overall similarities between the EGA sample structure and 2L FH compared
to other reference spectra suggest that EGA is also a Fe[III]oxyhydroxide, more similar
in structure to 2L FH that either goethite or scorodite. The strong feature noted by an
arrow at k = 7.8 A−1 tends to flatten as 3D structure is lost (Toner et al., 2008) (this
is bourne out in the significant decrease in this feature between goethite and 2L FH,
associated with a decrease in 3D order). This feature is further diminished in the EGA
sample compared to 2L FH, suggesting that the EGA may be less ordered than 2L FH.
4.3.7 Arsenic bonding structure - comparison to known arsenic-iron
complexes
In Figure 4.7, the k3-weighted As EXAFS spectrum for arsenic-laden EGA
generated at j = 1.1 mA/cm2 is compared to reference spectra for As[V] adsorbed onto
223
2 4 6 8 10 12 14
−10
010
20
30
40
50
k (A)
chi(k)*
k3
k(A )-1
χ(k
)*k3
mA/cm21.1
AsV ads on
2-Line Ferrihydrite
AsV ads on Goethite
Scorodite - FeAsO
mA/cm21.1
mA/cm21.1
4
Figure 4.7: Arsenic EXAFS spectrum (weighted by k3) for arsenic-laden EGA generatedat j = 1.1 mA/cm2 compared to reference spectra for As[V] adsorbed onto goethtite,As[V] adsorbed onto 2-line ferrihydrite (2L FH), and the mineral scorodite (FeAsO4).Arrows indicate spectral differences.
224
goethite, As[V] adsorbed onto 2-line ferrihydrite (2L FH), and the arsenic-iron mineral
scorodite (FeAsO4). The EGA sample is extremely similar in line shape, phase, and
amplitude characteristics both to As[V] adsorbed to goethite and As[V] adsorbed to 2L
FH. Although similar to scorodite, there are two key difference in line shape, indicated
by arrows. Differences in line shape are indicative of differences in the 3D arrangement
of atoms. The differences between scorodite and the EGA sample are severe enough to
imply that the majority of the arsenic in the sample is not forming scorodite. Based on
similarities with the top two reference spectra, the EGA sample is consistent with As[V]
adsorbed onto a Fe[III]oxyhydroxide mineral.
4.3.8 EGA iron structure - comparison between current densities
2 4 6 8 10
−5
05
10
15
k (A)
chi(k)*
k3
mA/cm2
mA/cm2
0.02
1.1
5.0
100
mA/cm2
mA/cm2
k(A )-1
χ(k
)*k
3
Figure 4.8: Iron EXAFS spectra (weighted by k3) for EGA generated at current densi-ties j = 100 mA/cm2, 5.0 mA/cm2, 1.1 mA/cm2, and 0.02 mA/cm2 (displayed top tobottom).
In Figure 4.8, the k3-weighted Fe EXAFS spectrum for EGA generated at
225
current densities j = 100 mA/cm2, 5.0 mA/cm2, 1.1 mA/cm2, and 0.02 mA/cm2 are
compared. The four spectra are all extremely similar in line shape, phase, and amplitude.
Some small differences exist in the amplitude near k = 10 A−1, but it is difficult to
attribute this to a change in structure because of the higher noise at high k, especially
in these spectra (see Section 4.2.4). In all other ways, these spectra are consistent with
the same 3D iron (hydr)oxide structure across all four current densities.
Significantly, this result is inconsistent with the hypothesis that the higher
arsenic removal capacity demonstrated at j = 0.02 mA/cm2 compared to j = 1.1 mA/cm2
(Section 3.3.4) is due to the increased presence of an iron (hydr)oxide with a higher
affinity for arsenic in the EGA. The EXAFS signal is averaged over all of the iron
(hydr)oxides present in EGA, and any significant differences in structure in the primary
component would appear in the EXFAS spectrum. An alternative hypothesis is discussed
more in Section 4.3.10 below.
Recall that from comparisons to iron reference compounds, EGA at j = 1.1 mA/cm2
was found to be consistent with a Fe[III]oxyhydroxide similar in structure to 2-line fer-
rihydrite (Section 4.3.6). Given the similarity in EGA structure across current density
shown here, EGA at j = 0.02 mA/cm2 is also consistent with an Fe[III]oxyhydroxide,
suggesting that EGA at 0.02 mA/cm2 is Fe[III] based rather than Fe[II] based. Similar
results were obtained from Fe XANES spectra above (Section 4.3.5). The provides fur-
ther evidence against the hypothesis suggested by polarization studies that EGA at j =
0.02 mA/cm2 is primarily Fe[II]-based (Section 3.3.10) .
226
4.3.9 Arsenic bonding structure - comparison between current densi-
ties
2 4 6 8 10 12 14
−10
010
20
30
EXAFS for 0.02,1.1,5.0,100 mA/cm^2
K (A^−1)
chi(k)*
k^3
mA/cm2
mA/cm2
0.02
1.1
5.0
100
mA/cm2
mA/cm2
k(A )-1
χ(k
)*k
3
Figure 4.9: Arsenic EXAFS spectra (weighted by k3) for EGA generated at currentdensities j = 100 mA/cm2, 5.0 mA/cm2, 1.1 mA/cm2, and 0.02 mA/cm2 (displayed topto bottom).
In Figure 4.9, the k3-weighted As EXAFS spectrum for EGA generated at
current densities j = 100 mA/cm2, 5.0 mA/cm2, 1.1 mA/cm2, and 0.02 mA/cm2 are
compared. As was the case for Fe EXAFS, the line shape, phase, and amplitude across
current densities is extremely similar. This suggests that the arsenic bonding struc-
ture, as well as the underlying structure of EGA bound to arsenic, is the same across
current densities. Given the comparison of EGA at j = 1.1 mA/cm2 to arsenic refer-
ence spectra above (Section 4.3.7), this suggests that the predominant bonding struc-
ture responsible for arsenic removal at all current densities is arsenic adsorption onto a
Fe[III]oxyhydroxide mineral.
227
4.3.10 Arsenic capture in the bulk solution compared to the electrode
surface
During ECAR treatment, iron (hydr)oxides are generated on the surface of
the electrode which may be very different in composition from the EGA generated in
the bulk solution. Arsenic may be captured on either the electrode surface or in the
bulk. As described in Section 4.2.4.3, several failed attempts were made to detect an
arsenic XAS signal from electrode surface layer samples scraped from the iron anode
immediately after ECAR treatment at j = 1.1 and 5.0 mA/cm2. This implies that most
of the arsenic removal in ECAR occurs in the bulk rather than on the electrode surface
at current densities j = 1.1 and 5.0 mA/cm2. This is consistent with results showing a
much higher As/Fe mass ratio in EGA sludge compared to electrode scrapings at j =
0.02mA/cm2 (Section 4.3.4). Interestingly however, the electrode scrapings at j = 0.02
mA/cm2 did show a strong arsenic XAS signal during the same XAS run and under the
same beam conditions as described above. The null signal at j = 1.1 and 5.0 mA/cm2
suggests that chemical analysis of these electrode scrapings would yield an even lower (if
existent) As/Fe ratio than the electrode scrapings at j = 0.02 mA/cm2.
The presence of arsenic on the electrode surface at j = 0.02 mA/cm2 compared
to no arsenic on the electrode surface at j = 1.1 and 5.0 mA/cm2 suggests an alternative
hypothesis for increased removal capacity at j = 0.02 mA/cm2. Perhaps at low current
densities, more arsenic is adsorbed to the electrode surface layer, increasing removal
capacity. This is further suggested by the fact that, during Matrix 2 batch tests (Chap-
ter 3), the electrode-surface-area to active-solution-volume ratio (A/V) was allowed to
228
vary to accommodate the factor of 5000 difference in current density between j = 0.02
and 100 mA/cm2. This resulted in a factor of 16 increase in electrode area relative to
the treatment volume for j = 0.02 mA/cm2 compared to j = 100 mA/cm2. Further
study is planned to explore this alternative hypothesis.
4.4 Chapter Summary
The reaction products of ECAR (known as EGA sludge) have been analyzed us-
ing SEM imaging, XRF mapping, chemical analysis, and X-ray absorption spectroscopy
(EXAFS and XANES). SEM images showed that EGA is very homogenous and made up
of rounded popcorn-like particles with a primary aggregate size of 50-100 nm. Arsenic
breakthrough experiments with increasing pore size showed that the cluster size of EGA
is as least 800 nm. XRF mapping of the EGA sludge confirmed homogeneity and showed
that arsenic is evenly captured across EGA.
Arsenic-laden EGA sludge is a waste product of ECAR and must be chemically
characterized for safe disposal. An initial chemical characterization (including only Fe
and As) showed that the waste from a representative ECAR treatment of synthetic
Bangladesh groundwater (j = 1.1 mA/cm2, q = 175 C/L, initial arsenic = 590 µg/L,
final arsenic = 5.0 µg/L, SBGW-2 recipe) contained 3300 ppm arsenic by mass with an
As/Fe mass ratio of 0.0076. Estimates of the expected amount of waste per treated liter
based on the amount of arsenic removed from solution at q = 150 C/L were a factor of
3 higher than the measured amount of waste produced. This indicates that more study
is needed for accurate waste estimates.
229
Fe XANES spectra for EGA generated at j = 0.02 - 100 mA/cm2 were consistent
with an iron oxidation state of 3+ at all current densities. This disproved the hypothesis
that EGA generated at j = 0.02 mA/cm2 is dominated by Fe[II]-based (hydr)oxides, as
suggested by polarization curves in Section 3.3.10.
Comparisons between Fe EXAFS spectra of EGA sludge generated at current
densities j = 0.02, 1.1, 5.0, and 100 mA/cm2 revealed no discernible difference in
iron (hydr)oxide structure with current density. Comparisons between arsenic EXAFS
spectra similarly revealed no discernible difference in the structure of the iron (hydr)oxide
bound to arsenic with current density. These results are inconsistent with the hypothesis
that differences in EGA composition with current density account for the differences in
arsenic removal capacity with current density measured in Chapter 3. An alternative
hypothesis was formulated that attributes the increase in arsenic removal capacity to
an increase in arsenic capture by the electrode surface layer at low current density.
This could potentially be accounted for by the increase in the electrode-surface-area to
active-solution-volume ratio (A/V) used in batch tests to accommodate the factor of 5000
difference in current density between j = 0.02 and 100 mA/cm2. Further experiments
are planned to test this hypothesis.
Based on Fe EXAFS spectral comparisons between EGA sludge and reference
iron (hydr)oxide compounds, EGA sludge was found to be consistent with a Fe[III]oxyhydroxide
similar in structure to 2-line ferrihydrite. This is consistent with the Fe XANES results
showing a 3+ oxidation state for iron.
Comparisons were made between As EXAFS spectra of EGA sludge and refer-
230
ence compounds, including arsenate adsorbed to 2-line ferrihydrite, arsenate adsorbed to
goethite, and the As-Fe mineral phase scorodite (FeAsO4). Based on strong similarities
in key spectral features such as line shape, phase, and amplitude, the arsenic bond-
ing structure responsible for arsenic removal in ECAR treatment was found to be very
consistent with arsenate adsorbed to a Fe[III]oxyhydroxide. Based on key differences
in line shape, the spectrum is inconsistent with scorodite. Further analysis is under-
way to determine the bonding structure from characteristic bond lengths determined by
shell-by-shell fits to the EXAFS spectra.
Combining these results, two important conclusions are reached - (1) there is
no significant difference in the arsenic removal mechanism of ECAR during operation
at different current densities (at least in the bulk solution), and (2) based on EXAFS
spectral comparisons, the arsenic removal mechanism in ECAR is consistent with ar-
senate adsorption onto a homogenous Fe[III]oxyhydroxide similar in structure to 2-line
ferrihydrite. Further analysis of the EXAFS spectra beyond the scope of this thesis will
allow for more concrete conclusions.
231
Chapter 5
ECAR Performance in Real
Bangladesh Groundwater
5.1 Introduction
It is difficult to model the diverse and complex chemistry state of real ground-
water. Although care was taken to build a synthetic groundwater using the average ion
concentrations found in Bangladesh, actual groundwater in Bangladesh covers a wide
range of concentrations of ions as well as composition of ions (i.e. different relative con-
centrations) (BGS, 2001). It is known, for example, that bicarbonate, phosphate, and
silicate can adversely affect arsenic adsorption onto iron (hydr)oxides (Section 2.4.6) and
that these ions exist in relatively high concentrations in the groundwater of Bangladesh
(Section 3.2.3). The presence of these ions together can interfere more than the sum
of each appearing alone (Meng et al., 2002), making it difficult to predict the relative
232
and absolute concentrations that will lead to a greater or lesser effect. In addition, the
ions in real groundwater have been interacting for much longer than in synthetic water,
and may have aged into compounds that cannot easily be mimicked in the synthetic
groundwater. Thus, it is essential to demonstrate the ability of ECAR to remove arsenic
from real arsenic-contaminated groundwater from tube well sources in Bangladesh.
In addition, observations of synthetic groundwater designed to mimic conditions
in Bangladesh (described in Section 3.2.3) showed almost complete oxidation of As[III]
to As[V] within 2 days. Iron (hydr)oxides are known to have a lower affinity for As[III],
which is also known to sorb at a slower rate (Bissen & Frimmel, 2003). During ECAR
treatment, there should be a net oxidation of As[III] to As[V] in the electrochemical
cell (Section 2.5.3), however the rate of this net reaction is not known, and there is
some suggestion that this may be the rate determining step for the residence time tr
(Section 3.3.9). It was also observed the that the residual arsenic after treatment is
>80% As[III] (Section 3.3.9). Thus it is prudent to explore if similar oxidation is seen
in real Bangladesh groundwater removed from a tube well. If oxidation does occur,
a waiting period before treatment could potentially increase the efficiency of arsenic
removal.
5.1.1 Issues of groundwater transport and storage
Due to laboratory and time constraints in Bangladesh, it was decided to trans-
port groundwater samples from tube wells in Bangladesh to the lab in Berkeley, CA
for ECAR testing. Transport occurred within three weeks of sample collection. The
233
transport of water inevitably changes certain water qualities, making it important to list
and understand the potential changes.
Based on observations of synthetic groundwater in the lab showing rapid oxi-
dation of As[III] to As[V], it is likely that most or all of the As[III] originally present
in real groundwater will be oxidized to As[V] by the time of testing. Gallagher et al.
(2004) reported changes in the ratio of As[III] to Astot in groundwater shipped overnight
without treatment. Batch tests of ECAR using synthetic groundwater showed that 80%
of the residual aqueous arsenic after treatment was As[III] (Section 3.2.1), implying that
ECAR generated iron (hydr)oxides have a higher affinity for As[V]. A reduction in the
amount of As[III] relative to As[V] could therefore artificially decrease the apparent
required dosage for complete arsenic removal in real Bangladesh groundwater.
Arsenic preservation problems in iron-rich water samples have been reported by
Aggett & Kriegman (1987), Borho & Wilderer (1997), Anderson et al. (1986), and Boyle
et al. (1998). Arsenic-contaminated groundwater in Bangladesh is known to contain
iron (average Fetot = 5.6 ± 5.9 mg/L according to National Survey data with Astot>10
µg/L in BGS (2001)). Iron(II) is abundant in well waters, and is slowly oxidized by
oxygen in air, forming iron[III]hydroxides which precipitate out of the water (Crisp &
Chowdhury, 2001). Calcium and bicarbonate are also abundant in well waters, including
those of Bangladesh. When exposed to air, carbon dioxide is lost and calcium carbonate
precipitates (Crisp & Chowdhury, 2001). Aqueous arsenic adsorbs to iron(III) hydroxide
and calcium carbonate, reducing the aqueous arsenic concentration in the water sample.
Mamtaz & Bache (2001) found that naturally occurring iron ≥ 1.1 mg/L in Bangladesh
234
groundwater could be used to reduce 100 µg/L As[III] to below 50 µg/L simply by
mixing the solution and allowing it to settle for 3 days. Similar processes will occur in
the samples as iron precipitates and settles, or sticks to the bottle sides - reducing the
arsenic concentration in the supernatant of the sample bottle from the original arsenic
concentration in the tube well water. Gallagher et al. (2004) reported losing 25-40%
of total arsenic during overnight transport in US groundwater samples containing iron
precipitate. Any loss in arsenic due to iron precipitation in the sample must be accounted
for by measuring the arsenic concentration in the supernatant immediately before testing.
This will not prevent arsenic loss, but will allow for an accurate understanding of the
arsenic removed due to ECAR treatment as opposed to iron precipitation during storage.
5.1.2 Research objectives
Specific research objectives for real Bangladesh groundwater included the fol-
lowing:
1. Demonstrate the ability of ECAR treatment to reduce arsenic levels from 100 - 500
µg/L in real Bangladesh groundwater to below either the Bangladesh maximum
allowable limit (50 µg/L) or ideally, the WHO recommended maximum limit (10
µg/L).
2. Measure the level of As[III] to As[V] oxidation in tube well water samples up to
seven days after collection.
3. Characterize the co-occurring solute composition of tube well water samples and
compare to regional and national survey data from BGS (2001).
235
5.2 Methods
5.2.1 Arsenic analysis and arsenic speciation
All reported arsenic concentrations were measured using ICPMS provided by
Curtis & Tompkins, Ltd (as described in Section 3.2.1) unless otherwise noted. Samples
for ICPMS were transported to Berkeley in plastic 15-mL vials that had been radiation
sterilized by the vendor (BD Biosciences). Preservation and digestion of samples occurred
in Berkeley, and in every case digested samples appeared clear and free of precipitate.
Quick Test arsenic measurements, as described in Section 3.2.1, were sometimes used in
the field and lab as a rough estimate of arsenic concentration. Separation of As[III] from
As[V] in the field was performed using arsenic speciation cartridges described in Section
3.2.1. Experiment specific procedures used in the field are described below.
5.2.2 Tube well water sample collection and storage
From March 23 to April 7, 2007, thirteen tube wells in Bangladesh were sam-
pled, twelve with high arsenic concentrations(Astot > 100 µg/L). Samples were collected
from five villages in Jhikargachha Upazila and Abhaynagar Upazila (both of Jessore dis-
trict in Khulna division) and one village from Sonargaon Upazila, just outside of Dhaka
(see Figure 5.1). Seven samples were later used in testing, listed in Table 5.1, including
at least one sample from each village vistited.
Villages containing tube wells with high levels of arsenic were identified by the
Bangladesh Rural Advancement Committee (BRAC), a Bangladeshi NGO, using the
final report of their 1999-2000 arsenic study (BRAC, 2000). During all field visits we
236
JhikargachhaTW 1, 3, 4
AbhaynagarTW 7, 10, 11
SonargoanTW 12
Figure 5.1: Map of Bangladesh highlighting areas visited. Map modified fromwww.homelandbangladesh.com.
237
were accompanied by at least one BRAC employee. In each village, BRAC employees,
community leaders, and local villagers helped identify households with contaminated
tube wells.
At each tube well, the initial pH, Dissolved Oxygen (DO), and arsenic concen-
tration were measured. The pH and DO were measured twice in the field and averaged,
using a portable pH and DO probe and meter (SympHony). The pH probe was discon-
nected and dried for travel to Bangladesh. Upon arrival, it was rinsed in tap water and
allowed to sit in electrolyte solution for 60 min. The pH probe was calibrated every other
day using pH 7.01 and pH 4.01 buffer solutions (Orion) which were stored in a cool dry
place. The DO probe was calibrated daily using a wet sponge (same procedure described
in 3.2.2). The arsenic concentration was estimated on site using the QuickTest. The goal
of the initial measurement was only to accept or reject the well as arsenic-contaminated,
thus the water was not initially diluted to be within the optimum measurement range
of the test, leading to initial results with high errors (up to 200%). For later samples,
20x dilutions were made using Jessore city tap water pre-filtered to remove particulates
(via an MSR hand-pumped water filter built for hiking) and subsequently determined
by ICPMS to contain < 1 µg/L arsenic.
If arsenic levels were sufficiently high according to the Quick Test (in most cases,
greater than 200 µg/L), two or three 1-liter tube well water samples were collected. All
samples were collected after pumping the tube well for five minutes in order to ensure that
the sample would be free from biological contamination and oxygenated water from the
238
tube well column1. All sample bottles (previously unused 1L high density polyethylene)
were filled to the brim and tightly capped to reduce exposure to oxygen. All sample
bottles were stored in the dark (to prevent algae growth and photo-oxidation) and not
opened again until the first testing was performed, either in Bangladesh (to monitor
As[III] to As[V] oxidation) or in Berkeley (ECAR treatment and measurement of co-
occuring solute concentrations). No acid or preservatives were added to the tube well
samples. Tube well samples are designated by TW x, where ”x” numbers the tube wells
in order of sample collection.
Aliquots for ICPMS measurement of the initial arsenic concentration were re-
moved from one sample bottle for each tube well within 0-9 hours of collection (samples
times listed in Table 5.1). In retrospect, initial water samples should have all been col-
lected at the tube well itself and preserved with nitric acid to avoid arsenic adsorption
to natural iron hydroxides precipitating in the sample bottle (see Section 5.1.1). It was
noted that within 1-2 hours of collection some of the plastic sample bottles containing
tube well water began to turn brown, with the exception of TW 12 from Sonargaon
Upazila. In addition, a small amount of brown precipitate began to form in the bottom
of the bottles, though TW 12 samples remained clear. The brown precipitate and bottle
discoloration indicate the precipitation of iron hydroxides, likely removing some of the
original arsenic present in the tube well water via adsorption and settling. Thus the
initial arsenic concentration listed in Table 5.1 is a lower limit of the arsenic present in
the tube well at the time of pumping.1Tube well depth is typically 40-80 m (Burgess & Ahmed, 2006).
239
Table 5.1: Tube well locations, initial arsenic concentration and time of initial aliquot,and list of tests performed on each tube well water sample.
Initial Time ofTube Well (TW) District Upazilaa Union Tests b As As Samplec
(µg/L) (Hrs)
1 Jessore Jhikargachha Godkhali I 667 9.03 Jessore Jhikargachha Godkhali A 444 7.54 Jessore Jhikargachha Godkhali A, E 333 6.57 Jessore Abhaynagar Prembug I 378 0.710 Jessore Abhaynagar Prembug E 378 2.711 Jessore Abhaynagar Prembug E 356 2.112 Narayanganj Sonargoan Amnipur I, A, E 600 0
a In Bangladesh, the Upazila is also known as P.S.b I = tested for a series of co-occuring solutes, A = used in As[III]/Astot monitoring, E = treated with ECAR.c Time between collection at the tube well and initial aliquot removal for initial arsenic ICPMS measurement.
5.2.3 Measurements of As[III]/Astot ratio over time
Three tube wells were chosen arbitrarily for As[III]/Astot monitoring over seven
days in the field - TW 3, 4, and 12. In each case, the As[III]concentration was measured
by filtering an aliquot of tube well water sample through a single disposable speciation
cartridge and storing the effluent for ICPMS arsenic measurement in Berkeley. For
TW 4, the tube well water was pre-filtered (via an MGR hand-pumped water filter)
to remove high levels of particulate matter before being filtered through the speciation
cartridge (as recommended by the manufacturer). The Astot concentration was measured
by removing an aliquot of tube well water sample and storing it unfiltered for ICPMS
arsenic measurement in Berkeley. Samples were not preserved or digested (via addition
of 10% HNO3) until 1-2 days before ICPMS testing in Berkeley.
For TW 3 and 4, the initial Astot aliquot was unfortunately removed from the
sample 6-8 hours after the As[III] aliquot was filtered. Due to iron precipitation observed
240
in the sample after 1-2 hours, the total aqueous arsenic is likely to have decreased during
the first 6-8 hours, allowing only an upper limit estimate of the initial As[III]/Astot ratio
in the tube well. For later measurements, aliquots for As[III] and Astot measurement
were removed at the same time. However, due to the iron precipitation, some As[III]
was likely removed while attached to iron hydroxide particles during filtration through
the speciation cartridge itself. The cartridge is filled with aluminosilicate sand to bind
to As[V]. However, if large iron hydroxide particles are passed, it effectively acts as a
sand filter. The Astot aliquot was not filtered through sand, again preventing a robust
comparison between the As[III] and Astot measured. A lower limit to the As[III]/Astot
ratio might have been determined for the 2-day and 4-day values based on the assump-
tion that As[III] could only decrease due to precipitate filtered out in the speciation
cartridge. However, large quantities of precipitate may have changed the ability of the
aluminosilicate to remove As[V] properly. Thus the data from these wells was thrown
out, and valuable lessons were learned about field speciation.
For TW 12, all aliquots to measure As[III] and Astot were removed at the same
time. Iron content appeared to be very low in TW 12 samples, evidenced by no visible
iron precipitate after 20 days and only a very slight browning of the sample bottle.
Measurements of Astot taken from the sample bottle demonstrate a decrease in Astot
of only 15% after 8 days of storage (Table 5.3), also suggesting low iron precipitation.
Finally, ICPMS measurement of the total iron concentration in 11 day old TW 12 samples
showed 0.1 mg/L (Table 5.4). Low iron precipitation will minimize the effect of As[III]
removal in speciation cartridges for TW 12. For comparison, the decrease in aqueous
241
Astot measured in the sample for TW 4, which contained visible iron precipitate, was
72% over 13 days (Table 5.3).
5.2.4 Chemical analysis for co-occuring solutes
One tube well water sample from each Upazila (identified in Table 5.1) was
tested for a series of co-occuring solutes (i.e. background ion composition) to determine
how representative the samples were of wells in the BGS. Samples were tested at Curtis &
Tomkins Laboratory (Berkeley, CA) for Cl – , NO –3 and SO 2 –
4 using ion chromatography
(EPA procedure 300.0), Ptot (EPA procedure 365.2), and Ca, Mg, Na, K, Fe, and Mn
using ICPMS (procedure EPA 6020). Between 11-19 days after collection, samples were
shaken in the original sample bottle before aliquots of 100-150 mL were removed for
solute testing. This is the same procedure used before aliquots were removed for ECAR
testing after aging approximately the same amount of time (described in Section 5.2.5),
so the measured solute concentrations are likely to represent the aqueous concentrations
of solutes present in aliquots tested with ECAR. Unfortunately, the tube well water
samples were not preserved properly to measure alkalinity or HCO –3 concentration at
the time of ECAR testing, and no equipment or lab was available and within our budget
capable of measuring Si or silicate at the time. In the future, it would be useful to obtain
pre-ECAR measurements of these solutes, given the detrimental effect they may have on
arsenic removal via ECAR (Section 2.4.6).
242
5.2.5 ECAR Treatment
Sealed 1-liter tube well water samples were opened in Berkeley 8-22 days after
collection (specific dates and tube wells listed in Table 5.3). Dissolved oxygen (DO) was
measured immediately after opening, with the change in DO from the time of collec-
tion presented in Table 5.3. A DO rise of <2.5 mg/L in most cases indicates minimal
oxygen contamination during transport and storage of samples. In the case of TW 11,
oxygen contamination was the highest (∆ DO = 4.16 mg/L) but the sample still had
not reached equilibrium with the atmosphere (DO = 7 - 9 mg/L). Sample pH changed
very little during transport and storage (average pH − pHcollection = −0.08, maximum
pH − pHcollection = −0.39). Four tube well water samples (TW 4, 10, 11 and 12) were
chosen to cover a range of initial arsenic concentrations as well as geographic locations
(listed in Table 5.1). Batch tests were performed on 850 mL aliquots of sample, decanted
from the sample bottle after vigorous shaking, with the exception of TW 4 which was
not shaken. ECAR procedures as described in Section 3.2.5 were used with the 850
mL-separated cathode cell (described in Section 3.2.4). Iron electrodes were rinsed in
dilute HCl followed by deionized water before each test. Current densities of 0.75 - 1.1
mA/cm2 were applied for a residence time of tr = 11 - 22 minutes (q = 86 - 175 C/L;
see Table 5.3 for the exact operating conditions used for each well). For comparison,
one batch test using SBGW-1 synthetic groundwater (Astot = 530 µg/L, As[III]/Astot =
0.58, pH = 7.41, DO = 1.27 mg/L, background ions described in Table 3.1) was dosed
concurrently using the same setup (j = 1.1 mA/cm2, tr = 22 min, q = 179.8 C/L).
Since only one ECAR test could be performed per 1-liter tube well water sample,
243
the applied dosage was purposefully higher than the minimal dosage required to reduce
600 µg/L to < 10 µg/L using SBGW-1 synthetic groundwater. Higher dosages were
used to remediate higher initial arsenic concentrations (dosage based on Quick Test
measurement of initial arsenic). The target current density for all tests was 1.1 mA/cm2,
however, lower current densities were used for water with a high resistivity so as not to
overload the Galvonostat at 1.1 mA/cm2. An alternative would have been to bring the
electrodes closer together, however an overload could not be discovered until treatment
was already underway, and a delay during dosage could have produced unknown side
effects that would be difficult to account for.
Dosed samples were mixed (as described in Section 3.2.5) with aliquots were
removed every 15 minutes starting at tM = 0 min followed by filtration (as described in
Section 3.2.5) and a Quick Test for arsenic. When the Quick Test measured Astot = 0
µg/L for two subsequent measurements, the mixing was stopped and the final filtrate was
sent for ICPMS arsenic testing. This resulted in tM = 15 - 30 min. Aliquots filtered after
tM = 0 min were sent for ICPMS testing as well. For TW 4, ICPMS measurements were
made at tM = 0, 15, 30, 45, and 60 min to verify the Quick Test results showing Astot=0
µg/L for successive measurements. Note that the filtration process takes 3-4 minutes
per aliquot, during which time the remaining solution is mixed as it flows through the
filter. The mixing time tM refers to the start of filtration. Thus the mixing time has an
error of ≈ +3 min in each case.
As noted in Section 3.2.1, each set of samples sent to C&T for ICPMS testing
was accompanied by a blank sample, containing only deionized (DI) water and 10%
244
HNO3 as a preservative. Usually, no arsenic was detected in blank samples. However,
the blank sample sent in with samples from TW 4 and TW 11 was reported to have
10 µg/L. A second blank sample was made using the same bottle of HNO3 used to
preserve TW 4 and TW 11 along with a blank sample using fresh HNO3. The fresh
HNO3 sample showed no arsenic (as expected), while the sample with acid used for
TW 4 and TW 11 was reported to have 18 µg/L, pointing to arsenic contamination
in the acid bottle. ECAR tests could not be repeated, since the majority of tube well
water was used up for a single batch test. In the end, a systematic error of 10 µg/L
was subtracted from TW 4 and TW 11 measurements, choosing the more conservative
arsenic value measured in the blanks. In addition to the systematic error, the ICPMS
measured value for TW 4 after tM = 0 min came out to be 170 µg/L - much higher than
the initial arsenic measured before the ECAR testing, Asinit = 103 µg/L, and too high
to be explained by a systematic error of 10 - 20 µg/L. Contamination during dosage is
possible from electrode use in previous batch tests, however contamination was not seen
elsewhere and all electrodes were washed in HCl before testing, making contamination
unlikely. Quick Test measurements at tM = 0 min reported 3 ±10 µg/L, consistent with
other tube wells, but inconsistent with the ICPMS measurement. Given the known acid
contamination affecting the ICPMS value, the Quick Test measurement was used for
this point only. ICPMS measurements for remaining points were <15 µg/L, and since
contamination could never artificially lower arsenic, these could not be affected by as
large of an anomaly as the tM = 0 min point. This experience led to new procedures for
adding acid to ICPMS samples, including redundancy to prevent contamination of the
245
Table 5.2: Astot, As[III], and As[III]/Astot ratio for TW 12 up to 7 days after collection.
Days sinceCollection Astot As[III] As[III]/Astot pH DO
(µg/L) (µg/L) (mg/L)
0 600 578 0.96 6.93 1.781.33 533 444 0.83 7.02 3.923.92 567 60 0.11 6.98 6.137.00 511 27 0.052 7.61 4.15
acid, as well as mandatory back-up samples kept without acid for every point.
In addition to ECAR batch tests performed on tube well samples, an aliquot of
one sample (TW 10) was simply filtered using a 0.1µm membrane (as described in Section
3.2.5) to obtain a measure of possible arsenic removal via complexation with natural iron
precipitates in the water. The TW 10 sample was filtered 12 days after collection and
the sample bottle was shaken before the aliquot was removed. Arsenic complexation to
iron precipitates in the water was the reason for the decrease in Astot between collection
and 12 days after collection (i.e. before ECAR treatment); thus the proper initial arsenic
concentration to measure removal via complexation to iron precipitates is the original
Astot measured in the sample as soon after collection as possible.
5.3 Results and Discussion
5.3.1 As[III] monitoring in samples of real groundwater
This section describes results and possible mechanisms of changing As[III] con-
tent of stored groundwater samples.
The ratio As[III]/Astot for TW 12 shows a clear decrease over 7 days, lowering
246
from 0.96 to 0.052 (Table 5.2). The total arsenic decreases by only 15% in the same
period (presumably due to complexation with settling iron precipitates in the water),
and cannot account for the 95% decrease in As[III] after 7 days, or the 90% decrease
after 4 days. As[III] is known to be slowly oxidized by dissolved O2, though the half-life
reported in the literature is inconsistent. Eary & Schramke (1990) measured an As[III]
half-life of 1 - 3 years in equilibrium with atmospheric oxygen - much slower than the
oxidation rate observed here. Kim & Nriagu (2000) report a half-life of 4 - 9 days for
As[III] in air-saturated groundwater with low iron content and pH 7.6 - 8.5. This is still
slower than the rate observed here, but is much closer. Two possible mechanisms for the
observed reduction in As[III] are described in the following.
The first scenario involves a measurement error. As[III] will form complexes
with natural iron precipitates in the water over time. ICPMS would detect this As[III]
in the total arsenic measurement, but the large precipitates and associated As[III] would
be filtered out in the As[III] speciation cartridge, underestimating the As[III] in the
sample. For the 95% decrease seen in As[III] to be accounted for by complexation to
iron precipitates, the removal capacity of the natural iron would have to be 0.551 mg/L
As[III] for 0.1 mg/L iron (measured from TW 12 after 9 days storage via ICPMS - see
Section 5.3.4), or equivalently, 55 mg-As[III]/mg-Fe. Measurements of As[III] removal
during coprecipitation with iron after addition of FeCl3 have been reported as 0.002 -
0.012 mg-As[III]/mg Fe at pH 7 (estimated from figure in Hering et al. (1996)). Thus
it is unlikely that this effect could account for the entire observed rapid reduction in
As[III].
247
The second scenario involves the oxidation of As[III] in conjunction with the
oxidation of Fe[II] over time. Hug & Leupin (2003) found that, parallel to the disappear-
ance of Fe[II] due to oxidation by dissolved O2, As[III] is partially oxidized to As[V] on a
time scale of hours. As[III] oxidation was attributed to reactive oxygen species that are
formed as intermediates during the reduction of O2 with Fe[II]. The same group found
that bicarbonate increases the fraction of oxidized As[III]. This could provide a possible
pathway for the rapid As[III] oxidation seen.
The rapid oxidation of As[III] observed in TW 12 is very similar to the rapid
oxidation observed in synthetic Bangladesh groundwater (Section 3.2.3). Figure 5.2
shows the percent As[III] (relative to initial concentration) remaining after 0 - 7 days for
three standard synthetic Bangladesh water batches (SBGW-1 type) compared to TW 12
(data for SBGW taken from Section 3.2.3). The similarity in the behavior of As[III]
oxidation between synthetic and real groundwater suggests that the same mechanism is
responsible in both cases. However, synthetic Bangladesh groundwater batches contained
no iron at all, requiring a different mechanism to be at play in each case or ruling out
the Fe[II] oxidation theory above. Unfortunately, the uncertainty in the As[III]/Astot
ratio measured (discussed in Section 5.2.3) makes it difficult to draw conclusions on the
basis of this data alone. A repeated and more careful experiment is required for further
speculation.
248
0 1 2 3 4 5 6 7
020
4060
8010
0
Water Batch/Sample Age (days)
Per
cent
Initi
al A
sIII
Rem
aini
ng
WB 38WB 43WB 45TW 12
Figure 5.2: Percent As[III] relative to initial concentration remaining with time for threesynthetic Bangladesh groundwater batches (WB; open symbols, dotted lines) and tubewell water sample 12 (TW 12; solid symbol, solid line). The initial As[III] concentrationfor WB 38 and 45 was not measured, so the value is assumed to be the average As[III]concentration for all measured water batches, As[III] = 304 ± 41 µg/L. This uncertaintyappears in all subsequent points, since the measured As[III] concentration must be di-vided by the initial As[III] concentration to calculate the percentage As[III] remaining.The uncertainty in As[III]/As[III]initial for TW 12 is estimated as the percentage decreasein Astot relative to the initial concentration plus the ICPMS measurement uncertaintyfor both As[III] and As[III]initial.
249
5.3.2 ECAR performance in real groundwater
ECAR treatment was able to reduce the arsenic concentration in real Bangladesh
groundwater to less than the WHO recommended limit (Astot ≤ 10 µg/L) in every case
(Figure 5.3), removing 97-99% of the total initial arsenic (Table 5.3). Pre-treatment
arsenic concentrations ranged from 93 - 510 µg/L, covering the concentrations found in
>96% of contaminated tube wells in Bangladesh (according the the BGS survey - Section
3.2.3). Coulombic dosage, residence time, and mixing time varied between tests (Table
5.3) to ensure removal to below the WHO limit. Dosages were purposely high to ensure
complete removal; thus the dosages listed in Table 5.3 should not be mistaken for the
minimum dosage required in real groundwater. Minimum dosage estimation will require
further controlled testing.
From Figure 5.3, we see that final arsenic concentrations for tube well sam-
ples treated with ECAR were comparable to, and slightly smaller than, final arsenic
concentrations for synthetic groundwater (SBGW-1). In fact, the arsenic concentration
immediately after Coulombic dosing (designated by tM = 0 min in Figure 5.3 and Ta-
ble 5.3) was already below the WHO limit for TW 4, 10, and 12 and quite close for
TW 11, while for synthetic groundwater the concentration is still above the less strin-
gent Bangladesh limit of 50 µg/L. Due to the different ECAR conditions, it is difficult
to make generalizations about the kinetics from this data alone, and further data and
discussion is included in Section 5.3.3 below.
It is clear from Figure 5.3 and Table 5.3 that coprecipitation using the natural
iron in the well water is insufficient on its own to reduce arsenic to below either the WHO
250
Tab
le5.
3:E
CA
Rop
erat
ing
cond
itio
nsan
dre
sult
sfo
rtr
eatm
ent
oftu
bew
ell
sam
ples
and
SBG
W-1
synt
heti
cgr
ound
wat
er.
Sam
ple
Init
ial
Init
ial
∆D
OC
urr
ent
Ch
arg
eIn
itia
lA
s res
%A
sR
es%
∆A
sS
am
ple
aA
geb
pH
cD
Oc
tran
sdD
ensi
tye
Den
sity
et C
Pe
t Me
Asc
t M=
0f
As r
esg
Rem
hA
s[II
I]/A
s tot
tran
sj
(Days)
(mg/L
)(m
g/L
)(m
A/cm
2)
(C/L
)(m
in)
(min
)(µ
g/L
)(µ
g/L
)(µ
g/L
)t M
=0i
TW
413
7.0
72.0
90.6
71.1
85.4
11
45
93
3k
2.2
98
0.2
572
TW
10
12
7.5
93.4
12.0
51.1
85.4
11
30
180
6.9
5.6
97
0.7
953
TW
11
22
7.3
85.4
64.1
61.1
,1.0
m161.2
22
15
280
12
2.2
99
0.0
022
TW
12
87.6
14.5
12.3
70.7
5174.7
33
15
510
4.0
3.9
99
...n
15
SB
GW
-1...
7.4
11.2
7...
1.1
170.8
22
30
530
62
7.8
99
0.8
0...
Filtration
Only
p
TW
10
12
7.5
93.4
12.0
5..
....
...
...
378q
144
...
62
...
...
aT
W=
Tu
be
Wel
lsa
mp
le,
SB
GW
-1=
Synth
etic
Ban
gla
des
hG
rou
nd
wate
r-
reci
pe
1(s
eeT
ab
le3.1
).b
Tim
eb
etw
een
collec
tion
at
tub
ew
ell
an
dE
CA
Rtr
eatm
ent.
cIn
itia
lre
fers
toim
med
iate
lyb
efore
EC
AR
trea
tmen
t.d
Ch
an
ge
inD
Oof
sam
ple
bet
wee
nco
llec
tion
at
the
tub
ew
ell
an
du
nse
alin
gin
Ber
kel
eyb
efore
EC
AR
trea
tmen
te
EC
AR
op
erati
ng
para
met
ers
use
dto
trea
tsa
mp
les
-cu
rren
td
ensi
ty(j
),ch
arg
ed
ensi
ty(q
),cu
rren
tp
roce
ssin
gti
me
(tC
P)
an
dm
ixin
gti
me
(tM
)are
des
crib
edb
riefl
yin
Sec
tion
1.4
.1an
din
more
det
ail
inS
ecti
on
2.5
.2.
fT
he
resi
du
al
ars
enic
con
centr
ati
on
att M
=0
min
(i.e
.b
efore
the
mix
ing
stage)
.g
Th
ere
sid
ual
ars
enic
con
centr
ati
on
aft
erth
em
ixin
gst
age
(i.e
.aft
erco
mp
lete
trea
tmen
t).
hP
erce
nt
ars
enic
rem
oved
aft
erco
mp
lete
EC
AR
trea
tmen
t.i
Rati
oof
As[
III]
toA
s tot
con
centr
ati
on
att M
=0
min
(i.e
.b
efore
the
mix
ing
stage)
.j
Th
isis
the
per
cent
dec
rease
inth
ears
enic
con
centr
ati
on
bet
wee
nco
llec
tion
at
the
tub
ew
ell
(see
Tab
le5.1
)an
dim
med
iate
lyb
efore
EC
AR
trea
tmen
t.k
Th
isis
aQ
uic
kT
est
resu
lt,
wit
hm
inim
um
erro
rof±
10µ
g/L
.N
oIC
PM
Sre
sult
isavailab
le-
see
Sec
tion
5.2
.5.
mC
urr
ent
den
sity
start
edat
1.1
mA/cm
2an
dw
as
low
ered
to1.0
mA/cm
2aft
er11
min
du
eto
agalv
on
ost
at
over
load
.n
As[
III]
data
att M
=0
min
isn
ot
availab
lefo
rth
istu
be
wel
l.p
Tu
be
wel
l10
filt
ered
thro
ugh
a0.1µ
mm
emb
ran
eon
ly-
no
EC
AR
trea
tmen
t.q
Th
isis
the
init
ial
ars
enic
con
centr
ati
on
mea
sure
d0.7
hrs
aft
erco
llec
tion
-se
eT
ab
le5.1
.
251
93
378
180
510 530
280
3
624.0126.9
144
7.83.92.25.62.20
100
200
300
400
500
TW 4 TW 10 TW 11 TW 12 SBGW Filtration only (TW
10)
Ars
en
ic C
on
cen
trati
on
(u
g/
L)
Initial As Immed after dosage (t_m =0 m) After complete ECAR (t_m > 15m)
Figure 5.3: Arsenic concentrations before and after ECAR treatment for groundwaterfrom four tube wells in Bangladesh along with SBGW-1 synthetic groundwater (SBGW),plus results from filtration alone (no ECAR) of TW 10 after 12 days of settling. Post-treatment concentrations are given for samples filtered immediately after Coulombicdosing (i.e. no mixing, of tM = 0 min) and after more than 15 min mixing (tM > 15min, see Table 5.3 for mixing times). Note that only Quick Test arsenic measurementsare available for TW 4, tM = 0 min; all other measurements are via ICPMS. Theinitial arsenic concentration for TW 10 and Filtered only TW 10 refers to the arsenicconcentration immediately before ECAR treatment and 0.7 hrs after collection from thetube well respectively.
252
or Bangladesh limits in TW 10. Filtration alone reduced the total arsenic from 378 µg/L
to 144 µg/L, a reduction of only 62%. It is difficult to compare this directly to ECAR,
since ECAR treatment on the same tube well used the residual arsenic left in the water
after some iron precipitation and settling occurred, resulting in a lower initial arsenic
concentration (180 µg/L compared to 378 µg/L). However, TW 10 demonstrates that
an initial settling period and filtration step before ECAR could lower the initial arsenic,
allowing ECAR treatment with a lower Coulombic dosage.
The reduction in the arsenic concentration between collection at the tube well
and immediately before ECAR treatment is most likely due to iron precipitation and
subsequent adsorption in the well water (Section 5.2.4). Thus the magnitude of the
decrease in arsenic can act as a relative measure of the amount of iron precipitation
that occurred in each sample. Table 5.3 lists the percent decrease in arsenic during
storage. As expected, TW 12, which appeared clear and free of precipitates, has the
lowest decrease (15% loss compared to 22 - 72% loss in other tube wells).
5.3.3 Mixing time in real groundwater compared to synthetic ground-
water
It is clear from Figure 5.3 that residual arsenic concentrations immediately
after Coulombic dosing (tM = 0 min) are lower for tube well samples than for synthetic
groundwater. In fact, residual arsenic for all tube well samples at tM = 0 min is below
the WHO limit of 10 µg/L. Significant time and money savings could result if the mixing
stage were found to be unnecessary for removal to < 10 µg/L in Bangladesh groundwater.
253
Further comparison of mixing time is allowed from TW 4 arsenic measurements at tM =
0, 15, 30, 45, and 60 min, shown in Figure 5.4. Corresponding measurements for synthetic
groundwater at tM = 0, 15, and 30 min are overlaid, showing a large discrepancy at
tM = 0 min and 15 min, and convergence on the same value at tM = 30 min. The
discrepancy is not due to the different residence time between TW 4 treatment and
synthetic groundwater treatment because, from Table 5.3, the residence time is lower
for TW 4 than for synthetic groundwater. Additional residence time for the synthetic
groundwater would only reduce the arsenic concentration at tM =0 min. However, it
is likely that the discrepancy is due to the large difference in initial arsenic present
in the synthetic groundwater compared to TW 4. Despite having twice the residence
time, over four times as much arsenic must be removed from the synthetic water. This
hypothesis would seem to be refuted by Figure 5.3, in which TW 12, with a comparable
initial arsenic concentration to the synthetic groundwater (Asinitial = 511 µg/L compared
to Asinitial = 533 µg/L) also has a significantly lower residual arsenic concentration at
tM = 0 min. However, from Table 5.3, we see that the residence time for TW 12 is
50% longer than that of synthetic groundwater. This additional time could account for
the discrepancy. TW 11 has the same residence time as the synthetic groundwater, and
again shows a much lower residual arsenic at tM = 0 min. However the initial arsenic for
TW 11 is lower compared to synthetic groundwater (Asinitial = 278 µg/L for TW 11),
which could account for the discrepancy. Thus it cannot be ruled out from these data
that the required mixing time to reach <10 µg/L for synthetic water is due to a higher
initial arsenic concentration.
254
0 10 20 30 40 50 60
010
2030
4050
6070
Mixing Time (min)
Ars
enic
Con
cent
ratio
n ((µµ
gL))
SBGW−1TW 4
Figure 5.4: Residual arsenic concentration as function of mixing time (tM ) for SBGW-1synthetic groundwater (Asinitial = 530 µg/L) and real groundwater from Tube Well 4(TW 4; Asinitial = 93 µg/L). Note that points overlap at tM = 30 min. All arsenicconcentrations were measured using ICPMS except for TW 4, tM= 0 min, which wasmeasured using the Quick Test (hence the higher error). Error in mixing time is + 3min for all points due to time of filtration.
255
While the higher initial arsenic concentration cannot be ruled out, a second
plausible explanation is the higher level of As[III] compared to As[V]. The synthetic
groundwater batch started with a measured As[III]/Astot ratio of 0.58. The tube well
samples are likely to have converted most of the original As[III] to As[V] by the time
of ECAR testing, based on observations of rapid (within 7 days) oxidation discussed in
Section 5.3.1. Initial As[III] aliquots were filtered through speciation cartidges, but the
shaking of sample bottles prior to removing the aliquots likely increased the bias due to
filtration of As[III] complexed to iron precipitates in the speciation cartridge (discussed
in Section 5.2.3). The post-treatment As[III] concentration was measured after filtration
through a 0.1µm membrane, and therefore should display little bias due to particulate
filtration from the speciation cartridge (some bias will still occur due to precipitation
that occurs after filtration through the 0.1µm membrane but before filtration through
the speciation cartidge - artificially reducing the measured As[III]/Astot ratio). The
measured ratio at tM = 0 min (in Table 5.3) for synthetic groundwater is 0.80, indicating
that most of the remaining arsenic is As[III]. The ratio for tube well samples varies from
0 - 0.79, and in all cases As[III] ≤ 4.4 µg/L. This is consistent with both the hypothesis
that very little As[III] was in the tube well samples to begin with and that high As[III]
content is the cause of the increased mixing time needed in the synthetic groundwater.
However, it is not enough to prove either theory.
These data suggest, but do not prove, that it might be possible to avoid or
minimize the mixing stage in real Bangladesh water by letting the water sit for 4 days
and allowing the As[III] to oxidize to As[V]. Further study is recommended.
256
5.3.4 Sample well comparison to BGS - co-occurring solutes
One groundwater sample from each region visited (designated by Bangladesh
Upazila) was analyzed for numerous co-occuring solutes after 11-19 days of storage to ap-
proximate the background composition of tube well samples immediately before ECAR
treatment. Tube well samples collected from the same region were close together, sug-
gesting similar geochemistry. The measured concentrations, shown in Table 5.4, can be
compared to average concentrations found in the BGS (discussed more fully in Section
3.2.3) for shallow tube wells from the same Upazila (496 Upazilas cover the 150 km2
area of Bangladesh). These local statistics from BGS roughly approximate the back-
ground composition of the tube well before transport and storage, though the averages
suffer from poor statistics (only a few wells exist in BGS from each region - 7 from
Jhikargachha and 4 each from Abhaynagar and Sonargaon). Statistics derived from the
full BGS database (covering 433 of the 496 Upazilas) as well as synthetic groundwater
recipes are shown for comparison. All statistics derived from BGS include the average
solute concentration for wells in the designated region with Astot ≥ 10 µg/L, along with
the standard deviation (a rough estimate of the spread, since the distributions are not
gaussian).
Comparison of sampled tube wells to regional BGS concentrations
Comparisons of measured tube well solute concentrations to BGS regional aver-
age concentrations can give a rough estimate of the effect of storage and transportation
on tube well solute composition. Inter-well variation within a region is approximated by
257
Tab
le5.
4:G
roun
dwat
erco
mpo
siti
onof
sele
cted
tube
wel
ls(T
W)
afte
r11
-19
days
ofst
orag
eco
mpa
red
toB
GS
stat
isti
csfr
omth
esa
me
regi
on.
Synt
heti
cgr
ound
wat
erpa
ram
eter
sus
edin
batc
hte
sts
(SB
GW
-1,2
)ar
ead
ded
for
com
pari
son
(fro
mT
able
3.1)
.B
GS
stat
isti
csin
clud
eal
lwel
lsav
aila
ble
inth
ede
sign
ated
regi
onw
ith
As tot≥
10µ
g/L
(sta
ndar
dde
viat
ion
give
nin
pare
nthe
sis)
.It
alic
sin
dica
tea
conc
entr
atio
nth
atfa
llsou
tsid
eof
the
rang
ede
fined
byth
eB
GS
regi
onal
aver
age
valu
e±
the
stan
dard
devi
atio
n.
Sou
rce
(age)
All
sam
ple
sC
l–S
O2–
4C
aM
gN
aK
Fe
Mn
PS
i%
BG
S<
Pf
%B
GS
<S
if%
BG
S<
P,
<S
ig
inre
gio
nmg/L
TW
1(1
9d
ays)
TW
1,3,
428
bdld
122
24
17
1.9
1.3
0.0
30.2
12
39h
BG
SJhik
arg
ach
ha
a...c
2.8
(5.5
)e116(2
4)
26(6
)18(7
)3.0
(0.9
)3.5
(1.4
)0.4
5(0
.35)
0.5
(0.3
)17.5
(2.0
)31
29
49
TW
7(1
8d
ays)
TW
7,
10,
11
7.6
bd
l81
29
26
2.2
0.9
01.6
0.1
743k
BG
SA
bhaynaga
ra
...
bd
l105(2
2)
27(6
)81(1
08)
3.1
(0.7
)3.7
(1.9
)0.7
(0.6
)0.5
(0.4
)18.3
(3.4
)31
41
52
TW
12
(11
days)
TW
12
14
bd
l52
14
21
4.0
0.1
01.4
0.5
31
55h
BG
SSonarg
aon
a...
0.1
7(0
.29)
120(7
5)
36(2
5)
35(2
3)
5.0
(1.8
)7.3
(5.3
)1.3
(0.7
)0.8
(0.3
)18.7
(4.5
)46
44
64
BG
Snationalb
80.8
(203)
4.6
(17.4
)66(5
3)
27(2
1)
94(1
83)
2.1
(3.1
)5.6
(5.9
)0.6
(0.8
)1.3
(1.5
)19.7
(5.1
)67
51
80
SB
GW
-1125
81
53
8134
00
00.0
10
50
0S
BG
W-2
125
861
8138
00
01.3
19.5
67
50
80
a7
wel
lsavailab
lein
BG
Sfo
rJh
ikarg
ach
ha,
4w
ells
each
availab
lefo
rA
bh
ayn
agar
an
dS
on
arg
aon
.b
Aver
ages
an
dst
an
dard
dev
iati
on
sfo
rall
BG
Sw
ells
wit
hA
s tot≥
10µ
g/L
inth
eS
pec
ial
Su
rvey
data
for
Cl–
(155
wel
ls)
an
dth
eF
ull
Su
rvey
data
for
all
oth
erio
ns
(1386
wel
ls).
cR
egio
nal
Cl–
con
centr
ati
on
sn
ot
availab
le(a
vailab
leon
lyin
spec
ial
surv
eyare
as)
.d
bd
l=
Bel
ow
Det
ecti
on
Lim
it.
Det
ecti
on
lim
itfo
rsu
lfate
is0.0
2m
g/L
.e
Note
that,
of
7w
ells
,tw
oh
ad
ver
yh
igh
sulf
ate
con
centr
ati
on
s(1
4.6
an
d4.8
mg/L
)an
d5
wer
eb
elow
the
det
ecti
on
lim
itof
0.0
2m
g/L
.f
Th
ep
erce
nta
ge
of
BG
Sw
ells
inth
eF
ull
Su
rvey
wit
hA
s tot≥
10µ
g/L
wit
hco
nce
ntr
ati
on
sle
ssth
an
Por
Si
gT
he
per
centa
ge
of
BG
Sw
ells
inth
eF
ull
Su
rvey
wit
hA
s tot≥
10µ
g/L
wit
hP
con
centr
ati
on
sle
ssth
an
Pan
d/or
Si
con
centr
ati
on
sle
ssth
an
Si
(i.e
.th
isex
clu
des
all
wel
lsth
at
have
bothP
BG
S>
Pan
dSi B
GS>
Si)
.h
Ass
um
esS
ico
nce
ntr
ati
on
iseq
ual
tore
gio
nal
aver
age.
258
the standard deviation, which is admittedly uncertain due to a lack of gaussian distri-
butions and low counting statistics.
Not surprisingly, Table 5.4 indicates that all aged TW samples tested contain
below-average Fe concentrations for their region - this is likely an indicator of how much
Fe has precipitated out of solution and/or stuck to the sides of the sample bottle. Oddly,
TW 12, which contained no visible iron precipitate, has the lowest Fe concentration
compared to the highest regional average values in BGS. This suggests that TW 12
contained a low Fe concentration to begin with. Mamtaz & Bache (2001) reports that
natural iron in Bangladesh groundwater of a concentration Fe[III] = 66 ∗As1.75 where
Fe[III], As are in mg/L is sufficient to reduce arsenic concentrations to below 50µg/L
after shaking and 3 days settling. For TW 10 from Abhaynagar, initial arsenic at the
tube well (2.7 hours after collection - Table 5.1) was 378 µg/L, requiring 12 mg/L of
iron(III) to reduce arsenic to <50 µg/L using this method. Neither the measured value
of iron in nearby TW 7 nor the regional average in Abhaynagar are large enough to
support removal via natural arsenic alone, explaining the low 62% removal seen using
settling and filtration alone (Figure 5.3) for TW 10.
Other concentrations in tube well samples that are low or high compared to re-
gional averages include the Manganese (Mn) concentration for TW 1 and 7. Manganese
is another metal that can form insoluble hydroxides upon exposure to oxygen, perhaps
precipitating out with the iron. This could explain the lower than average concentra-
tion measured in TW 1. The presence of synthetic Mn oxyhydroxides was found to aid
the rapid oxidation of As[III] (Devitre et al. (1991)) and could perhaps be one cause
259
of extremely low As[III] concentrations found in TW 12 after 4 days (Section 5.3.1).
However, the Mn level measured in TW 12 is right within the range expected for the
region, suggesting little Mn has precipitated out and settled, though some could have
been disturbed in transit. Synthetic groundwater contained no Mn to begin with, imply-
ing a different effect would have to be the cause of rapid oxidation in real groundwater
compared to synthetic (which is possible). The Mn concentration measured for TW 7 is
above both the regional and national averages, however, high Mn is not likely to affect
adversely ECAR treatment.
Sulfate is lower than the regional average for TW 1. However, in this case, it
does not imply that the sulfate concentration is abnormally low since 5 of the 7 wells
used in the average contained no sulfate (i.e. levels were below detection limit of 0.02
mg/L). The remaining 2 wells contained abnormally high levels of sulfate, inflating the
average.
TW 7 has low K concentrations compared to the regional average, but the
concentration is right in the range of the national average. Na is also on the low side for
this well, though it is within the regional range. Low Na and K could slightly decrease
the conductivity of this water relative to average. This could help explain the low
conductivity displayed by the overloading galvonostat during ECAR Coulombic dosing
of TW 11 (Section 5.2.5). However, no overloading occurred during ECAR treatment
of TW 4, likely to have even lower Na and K concentration based on the regional and
measured TW 1 values (though higher Ca concentrations could make up for the low Na).
Phosphorous is within range, but on the low side for all aged tube well samples.
260
This could be due to either a relatively low phosphorous concentration in the tube well at
the time of collection, or adsorption of phosphate onto iron hydroxide precipitates in the
sample bottle. In either case, phosphorous at the time of ECAR testing is representative
of the phosphorous levels in the region.
While silicate was not measured in the aged tube well samples, the regional
average concentrations of silica fall within a tight range for all available wells in the
region. This suggests that the original silica concentration in our tube well samples was
within this range as well, though little is known about how much may have been removed
due to iron precipitation during aging.
All other co-occuring solutes are within the range defined by the average BGS
regional value ±the standard deviation.
Comparison of sampled tube wells to national BGS concentrations
A comparison of aged tube well solute concentrations and regional BGS averages
to the national BGS averages can help characterize how representative our tube well
samples are to tube wells across Bangladesh, specifically in terms of interfering ions (see
Section 2.4.6). The key solutes that can be detrimental to arsenic removal are phosphate,
silicate, and bicarbonate and to some extent sulfate (see Section 2.4.6).
Bicarbonate representativeness cannot be characterized from the current data,
both because it was not measured in our samples and because it is not available in the
Full Survey data of the BGS. Phosphorous concentrations in both the aged tube well
samples and BGS regional averages are 40-90 % lower than the national average. In the
261
BGS Full Survey, only 31% of wells tested (with Astot ≥ 10 µg/L) have phosphorous levels
below 0.5 mg/L, the highest value measured in the three aged samples. BGS regional
average silica concentrations are 5-11% less than the national average, but well within
one standard deviation. The lowest regional silica average, occurring in Jhikargachha,
is greater than 29% of the BGS Full Survey wells (with Astot ≥ 10 µg/L), putting it
firmly on the low side. Similar to the phosphorous concentration in the aged tube well
samples, the initial silica is likely to have decreased further before ECAR treatment due
to iron precipitation. Given these results, it is likely that the tube well samples used in
ECAR treatment tests contained relatively low levels of key solutes such as phosphate
and silicate compared to national averages.
Phosphate and silicate present together are known to have a more detrimental
effect on arsenic removal than the combination of effects from either alone (Section 2.4.6).
In Table 5.4, the percentage of BGS wells in the Full Survey with less P, less Si, or less
P and Si combined than the measured values is listed (for aged TW samples, the Si
concentration is assumed to be the regional average). Essentially, this estimates the
percentage of BGS wells for which the combined detrimental effect of phosphate and
silicate is likely to be worse than the effect seen in sample tube wells. By this metric,
sample tube wells represent 39 - 55% of the tube wells in Bangladesh in terms of combined
phosphate and silicate interference.
From Section 2.4.6, sulfate has been found to have little to no effect on As[V]
adsorption to ferrihydrite. As[III] adsorption was found to decrease a few percent for pH
<7 for S(VI)/As ratios of 10:1 and 50:1. Assuming the national average concentration of
262
sulfate (4.6 mg/L SO4, 1.54 mg/L S(VI)) and the initial arsenic concentrations of aged
tube well samples used in ECAR experiments (Table 5.3), the ratio of S(VI)/As ranges
from 17:1 to 3:1, within the range of measured slight decrease in As[III] adsorption. The
measured sulfate levels for aged tube well samples and BGS regional averages (Table 5.4)
are therefore low with respect to possible detrimental effects due to sulfate that might
occur in other regions. However, the effect on As[III] adsorption near pH 7 is very small
(1%) even for 50:1 S(VI)/As, and not likely to effect ECAR treatment significantly.
In fact, higher sulfate levels could aid ECAR efficiency due to the increase in water
conductivity.
5.4 Chapter Summary
Real groundwater from a tube well in Bangladesh was collected from the Sonar-
gaon Upazila (initial Astot = 600 µg/L) and the As[III]/Astot ratio was monitored up to
7 days after collection. A 95% decrease in the initial As[III] concentration was observed
over 7 days, with a 90% decrease after only 4 days. This is similar to the rapid oxidation
seen in synthetic Bangladesh groundwater over a similar time scale. In real groundwater,
95% oxidation took slightly longer than in synthetic groundwater (7 days compared to
3 days). Observed oxidation is too rapid to be due to atmospheric oxygen alone. More
careful measurements are required to determine the cause of rapid oxidation.
Real groundwater from Bangladesh was collected from high-arsenic (100 - 500
µg/L) tube wells in three regions and treated using ECAR. In every case, the final
arsenic concentration was reduced to below the WHO recommended limit of 10 µg/L.
263
Simple filtration of one tube well sample (i.e. with no ECAR) was only able to reduce
the arsenic concentration from 378 µg/L to 144 µg/L after 12 days of settling due to
complexation with naturally occurring iron precipitates in the water, indicating that
further treatment is necessary for this region. ECAR treatment in real groundwater
was was able to reduce the arsenic concentration to <12 µg/L in every case with no
mixing time, compared to synthetic groundwater which required at least 30 minutes of
mixing. This could be due to the low initial As[III]/Astot ratio of tested tube well water
compared to synthetic groundwater, suggesting that oxidation of As[III] prior to ECAR
treatment could remove the need for a mixing step. However, different residence times
used for the ECAR treatment of tube well samples (to control dosage) make it difficult to
generalize about the mixing time relative to synthetic groundwater; further experiments
are recommended with residence time held constant.
Concentrations of several co-occuring solutes were measured in representative
tube well water samples and compared to regional and national averages in the BGS.
ECAR-tested tube wells were found to have low initial phosphate concentrations com-
pared to regional and national tube wells (possibly due to complexation with iron pre-
cipitates and settling out of solution). Silica concentrations for BGS tube wells in each
region were within range of national values. Assuming the BGS regional values for silica
and all P and Si in the form of phosphate and silicate, measured tube wells contained
more phosphate and silicate in combination than 39-55% of tube wells in Bangladesh.
There was some suggestion through low sulfate, potassium, and sodium concentrations
that conductivity might be lower than average in the tube wells tested.
264
Valuable information was also gained on methods and procedures for sampling
and testing arsenic-contaminated groundwater in Bangladesh. The importance of on-site
ECAR testing was highlighted by issues of natural iron precipitation in the well water.
Although the iron precipitation is likely to aid arsenic removal through adsorption of
some arsenic, phosphate, and silicate, it adversely affects predictions of ECAR perfor-
mance in groundwater straight from the tube well. In addition, valuable information
was learned about the use and limitations of disposable speciation cartidges as a way to
measure As[III] in water with particulate matter.
265
Chapter 6
Conclusions
6.1 Summary of results
Electrochemical arsenic remediation (ECAR) using iron has been proposed as a
potentially viable, low-cost, and effective technology for community scale arsenic remedi-
ation in rural Bangladesh. Two bench-top electrochemical cells have been designed and
incorporated into a small-scale ECAR batch treatment procedure. Synthetic Bangladesh
groundwater (SBGW) has been developed to mimic key environmental properties that
could affect arsenic removal in Bangladesh, such as high levels of phosphate and silicate.
Arsenic removal capacity - defined as the average amount of arsenic removed per coulomb
between 600 µg/L and the WHO recommended maximum arsenic limit of 10 µg/L - has
been proposed as a performance metric directly related to the operating costs of ECAR
treatment. The arsenic removal capability to reach the WHO limit in the groundwater
environment of Bangladesh and total treatment time have been proposed as additional
266
key performance metrics for assessing ECAR.
The arsenic removal capability of ECAR has been demonstrated by reducing
initial concentrations of 550 - 580 µg/L arsenic (including equal amounts of As[III] and
As[V]) to below 10 µg/L in synthetic Bangladesh groundwater.
The effect of ECAR operating parameters - including current density, j, charge
density, q, mixing time, tM , and current processing time, tCP - on arsenic removal capac-
ity and total treatment time has been explored, resulting in the following observations
and measurements:
1. Increasing the duration of the post-electrolysis mixing stage in ECAR, controlled
by tM , can be used to increase arsenic removal capacity at the expense of increasing
the treatment time.
2. Arsenic removal during the mixing stage is biphasic, with a period of rapid arsenic
reduction between 10-40 minutes long (depending in part on tCP ) followed by slow
reduction.
3. For a given value of j, tM , and tCP , the total charge density passed into solution
determines the extent of arsenic removal, with increasing charge density leading to
increased removal.
4. Arsenic removal capacity is approximately constant within certain ranges of current
density, but highly variable between current density ranges. In order of decreasing
removal capacity, the pattern discernible from batch tests is: 0.02 mA/cm2 >
0.07 mA/cm2 > 0.30 mA/cm2 - 1.1 mA/cm2 > 5.0 mA/cm2 - 100 mA/cm2.
267
5. For a given value of q and j, Arsenic removal capacity increases with increasing
tCP at the expense of increasing the total treatment time.
6. No single set of operating parameters is optimal in terms of both arsenic removal
capacity and treatment time.
7. The highest arsenic removal capacity for reduction of arsenic to the WHO limit
in SBGW with a total treatment time of less than 2.5 hrs was 3.7 µg-Asrem/C,
achieved at j = 1.1 mA/cm2, q=150 C/L, tCP = 69 min, tM = 60 min.
8. The highest arsenic removal capacity overall for reduction of arsenic to the WHO
limit in SBGW was 22 µg-Asrem/C, achieved at j = 0.02 mA/cm2, q = 25 C/L,
tCP = 416 min (∼ 7 hrs), tM = 60 min.
This assessment differs from previous assessments of operating parameter influence on
arsenic removal via electrocoagulation (Kumar et al., 2004; Gomes et al., 2007) in the
following key points: (1) current density was found to have an effect on arsenic removal
capacity (though no effect was seen within the current density ranges studied by Kumar
et al. (2004) and Gomes et al. (2007)), (2) tCP was found to have an effect on arsenic
removal capacity that is independent of the total charge density and current density, and
(3) the addition of a post-electrolysis mixing stage controlled by the parameter tM was
added to allow for further increases in the arsenic removal capacity.
Post-synthesis ECAR-generated iron (hydr)oxides adsorbent, or PS-EGA ad-
sorbent, was created by running the ECAR dosing process in synthetic Bangladesh
groundwater containing no arsenic. Freshly made PS-EGA was able to reduce initial
268
concentrations of ∼ 600 µg/L (including As[III] and As[V] in a 1:1 ratio) to 37 µg/L
after 60 minutes of mixing and 22 µg/L after 48 hours of contact. ECAR treatment
under comparable conditions was able to reduce arsenic to 10 µg/L after 60 minutes of
mixing. The low final arsenic concentration using PS-EGA suggests that some of the
initial 300 µg/L of As[III] was able to complex with PS-EGA directly, without electro-
chemical oxidation. Significantly however, the adsorbent was not able to reduce total
arsenic levels to below the WHO limit (10 µg/L), indicating that the in-situ generation
of EGA during ECAR is significant to its arsenic removal capability.
Arsenic removal capacity for PS-EGA (as µg-Asrem/mg-Fe 3+) was 33% lower
after 60 minutes of contact if the PS-EGA was allowed to age 60 minutes before coming
into contact with arsenic. After 48 hours of contact, removal capacity was only 13%
lower, however, the minimum arsenic concentration of achieved was 100 µg/L. This
indicates that removal capacity of EGA is highly sensitive to aging. This is likely due
to the formation of aggregates. Once aggregates form, arsenic oxyanions must diffuse
into them to find available sites. It is possible that the adsorption capacity of aged-
PS-EGA could catch up with that of fresh-PS-EGA given enough time for diffusion
into aggregates to occur. However, given the constraints on treatment time proposed in
Chapter 1, attempting to form PS-EGA at a central location and shipping the adsorbent
to satellite treatment sites in place of on-site ECAR treatment is not advised.
The arsenic removal capacity of ECAR for As[III] appears to be approximately
3 times lower than the removal capacity for As[V]. This is consistent with both (1) a lower
affinity of EGA for As[III] compared to As[V] or (2) slow oxidation of As[III] compared
269
to the rate of As[V] complexation, requiring more current processing time rather than
more charge.
Freshly made PS-EGA (defined in previous paragraphs) was able to remove 90%
of added As[III] after 60 minutes of contact in SBGW containing competitive ions such
as phosphate, silicate, bicarbonate, and As[V]. No electrodes were present or external
voltage applied during the contact time, preventing electrochemical oxidation of As[III].
This implies that EGA has some affinity for As[III] on the timescale of ECAR treatment,
even in the presence of competing ions. However, ECAR is still capable of reducing
As[III] concentrations lower than adsorption onto PS-EGA even after 48 hours of contact.
These results suggest that (1) direct complexation of As[III] onto EGA may account for
some portion of arsenic removal in ECAR and (2) some additional process (possibly
electrochemical oxidation) is responsible for the reduction of As[III] down to the lowest
achievable levels near the WHO limit.
Slow scan transient voltammetry was performed on the iron electrode at ν =
0.1 mV/s in 0.1 M KClO4 (plain perchlorate) solution and 0.1 M KClO4 plus SBGW. A
significant current drop in the SBGW scan relative to plain perchlorate solution suggests
that species in the SBGW interfere with the dissolution processes on the iron electrode.
This resulting increase in the overpotential required to reach a given current density in
SBGW will add to the energy consumption of ECAR treatment. Since the overpotential
is less severe at higher current densities, this could give high current densities some
advantage over low current densities in terms of energy consumption. This could help
offset disadvantages due to higher operating current.
270
Electrode passivation, a common problem during electrocoagulation with alu-
minum electrodes, was not seen within the range of the polarization scan (up to current
density j = 10 mA/cm2). The lack of passivation in iron over the current density region
that is most effective for arsenic removal is promising in terms of electrode longevity,
maintenance requirements, and the ability to operate in monopolar mode.
The polarization scan revealed that the electrode potential over the entire range
of scanned current density (i.e. j ≤ 10 mA/cm2) is well below the thermodynamic
equilibrium potential for oxygen evolution. This rules out oxygen evolution as a cause
of the decrease in arsenic removal capacity seen at higher current density.
Initial chemical characterization (including only Fe and As) of arsenic-laden
EGA showed that the waste from a representative ECAR treatment of synthetic Bangladesh
groundwater (j = 1.1 mA/cm2, q = 175 C/L, initial arsenic = 590 µg/L, final arsenic
= 5.0 µg/L, SBGW-2 recipe) contained 3300 ppm arsenic by mass with an As/Fe mass
ratio of 0.0076. Measured waste quantity was 0.33 mg waste per liter of treated water.
Estimates of the expected amount of waste per treated liter based on the amount of ar-
senic removed from solution at q = 150 C/L were a factor of 3 higher than the measured
amount of waste produced.
SEM images showed that EGA is very homogenous and made up of rounded
popcorn-like particles with a primary aggregate size of 50-100 nm. Arsenic breakthrough
experiments with increasing pore size showed that the cluster size of EGA is as least
800 nm. XRF mapping of the EGA sludge confirmed homogeneity and showed that
arsenic is evenly captured across EGA.
271
A direct of comparison of 0.1 µmfiltration to sedimentation (i.e. quiescent set-
tling followed by decantation) as a means to separate EGA from potable water revealed
that approximately three days of settling is required to achieve comparable arsenic re-
moval results. The long required settling time is likely due to the small particle size
of EGA. An alternative low-cost method of EGA separation compatible with the small
particle size will have to be explored.
Fe XANES spectra for EGA generated at j = 0.02 - 100 mA/cm2 were consistent
with an iron oxidation state of 3+ at all current densities. This disproved the hypothesis
that EGA generated at j = 0.02 mA/cm2 is dominated by Fe[II]-based (hydr)oxides, as
suggested by polarization curves in Section 3.3.10.
Comparisons between Fe EXAFS spectra of EGA sludge generated at current
densities j = 0.02, 1.1, 5.0, and 100 mA/cm2 revealed no discernible difference in
iron (hydr)oxide structure with current density. Comparisons between arsenic EXAFS
spectra similarly revealed no discernible difference in the structure of the iron (hydr)oxide
bound to arsenic with current density. These results are inconsistent with the hypothesis
that differences in EGA composition with current density account for the differences in
arsenic removal capacity with current density measured in Chapter 3. An alternative
hypothesis was formulated that attributes the increase in arsenic removal capacity to
an increase in arsenic capture by the electrode surface layer at low current density.
This could potentially be accounted for by the increase in the electrode-surface-area to
active-solution-volume ratio (A/V) used in batch tests to accommodate the factor of 5000
difference in current density between j = 0.02 and 100 mA/cm2. Further experiments
272
are planned to test this hypothesis.
Based on Fe EXAFS spectral comparisons between EGA sludge and reference
iron (hydr)oxide compounds, EGA sludge was found to be consistent with a Fe[III]oxyhydroxide
similar in structure to 2-line ferrihydrite. This is consistent with the Fe XANES results
showing a 3+ oxidation state for iron.
The arsenic bonding structure responsible for arsenic removal in ECAR treat-
ment was found to be very consistent with arsenate adsorbed to a Fe[III]oxyhydroxide.
Based on key differences in line shape, the spectrum is inconsistent with scorodite. Fur-
ther analysis is underway to determine the bonding structure from characteristic bond
lengths determined by shell-by-shell fits to the EXAFS spectra.
Combining these results, two important conclusions are reached - (1) there is
no significant difference in the arsenic removal mechanism of ECAR during operation
at different current densities (at least in the bulk solution), and (2) based on EXAFS
spectral comparisons, the arsenic removal mechanism in ECAR is consistent with ar-
senate adsorption onto a homogenous Fe[III]oxyhydroxide similar in structure to 2-line
ferrihydrite. Further analysis of the EXAFS spectra beyond the scope of this thesis will
allow for more concrete conclusions.
Real groundwater from a tube well in Bangladesh was collected from the Sonar-
gaon Upazila (initial Astot = 600 µg/L) and the As[III]/Astot ratio was monitored up to
7 days after collection. A 95% decrease in the initial As[III] concentration was observed
over 7 days, with a 90% decrease after only 4 days. This is similar to the rapid oxidation
seen in synthetic Bangladesh groundwater over a similar time scale. In real groundwater,
273
95% oxidation took slightly longer than in synthetic groundwater (7 days compared to
3 days). Observed oxidation is too rapid to be due to atmospheric oxygen alone.
Real groundwater from Bangladesh was collected from high-arsenic (100 - 500
µg/L) tube wells in three regions and treated using ECAR. In every case, the final
arsenic concentration was reduced to below the WHO recommended limit of 10 µg/L.
Simple filtration of one tube well sample (i.e. with no ECAR) was only able to reduce
the arsenic concentration from 378 µg/L to 144 µg/L after 12 days of settling due to
complexation with naturally occurring iron precipitates in the water, indicating that
further treatment is necessary for this region. ECAR treatment in real groundwater was
was able to reduce the arsenic concentration to <12 µg/L in every case with no mixing
time, compared to synthetic groundwater which required at least 30 minutes of mixing.
This could be due to the low initial As[III]/Astot ratio of tested tube well water compared
to synthetic groundwater, suggesting that oxidation of As[III] prior to ECAR treatment
could remove the need for a mixing step. However, different current processing times
used for the ECAR treatment of tube well samples (to control dosage) make it difficult to
generalize about the mixing time relative to synthetic groundwater; further experiments
are recommended with current processing time held constant.
Concentrations of several co-occuring solutes were measured in representative
tube well water samples and compared to regional and national averages in the BGS.
ECAR-tested tube wells were found to have low initial phosphate concentrations com-
pared to regional and national tube wells (possibly due to complexation with iron pre-
cipitates and settling out of solution). Silica concentrations for BGS tube wells in each
274
region were within range of national values. Assuming the BGS regional values for silica
and all P and Si in the form of phosphate and silicate, measured tube wells contained
more phosphate and silicate in combination than 39-55% of tube wells in Bangladesh.
There was some suggestion through low sulfate, potassium, and sodium concentrations
that conductivity might be lower than average in the tube wells tested.
Valuable information was also gained on methods and procedures for sampling
and testing arsenic-contaminated groundwater in Bangladesh. The importance of on-site
ECAR testing was highlighted by issues of natural iron precipitation in the well water.
Although the iron precipitation is likely to aid arsenic removal through adsorption of
some arsenic, phosphate, and silicate, it adversely affects predictions of ECAR perfor-
mance in groundwater straight from the tube well. In addition, valuable information
was learned about the use and limitations of disposable speciation cartidges as a way to
measure As[III] in water with particulate matter.
6.2 Future Work
A prototype continuous flow device based on ECAR was design and fabricated
by a team of graduate students (under the direction of Ashok Gadgil and myself) in
Spring 2008. The prototype was found to successfully reduce 600 µg/L As to below the
WHO limit in synthetic Bangladesh groundwater (data to be published at a later date).
The prototype was recently operated in Bangladesh and arsenic-contaminated region of
Cambodia, achieving very promising results. It is currently being prepared for more
rigorous field-testing in Bangladesh.
275
Bibliography
Abbgy, A., Kelly, T., Lawrie, C., & Riggs, K. (2002). Quick Arsenic Test kit. Tech. rep.,Battelle.
Abdullah, M., Shiyu, Z., & Mosgren, K. (1995). Arsenic and selenium species in the oxicand anoxic waters of the Oslof-jord, Norway. March Pollution Bulletin, 31 , 116–126.
Abuzaid, N. S., Bukhari, A. A., & Al-Hamouz, Z. M. (2002). Ground water coagulationusing soluble stainless steel electrodes. Advances in Environmental Research, 6 (3),325–333.
Aggett, J., & Kriegman, M. (1987). Preservation of arsenic (III) and arsenic (V) insamples of sediment interstitial water. The Analyst , 112 , 153–157.
Ahmad, J., Goldar, B., Misra, S., & Jakaruya, M. (2003). Willingness to pay for arsenic-free, safe drinking water in Bangladesh. Tech. rep., The World Bank.
Ahmed Basha, C., Bhadrinarayana, N. S., Anantharaman, N., & Meera Sheriffa Begum,K. M. (2008). Heavy metal removal from copper smelting effluent using electrochemicalcylindrical flow reactor. Journal of Hazardous Materials, 152 (1), 71–78.
Ahsan, H., Chen, Y., Parvez, F., Argos, M., Hussain, A. I., Momotaj, H., Levy, D., vanGeen, A., Howe, G., & Graziano, J. (2006). Health effects of arsenic longitudinal study(HEALS): Description of a multidisciplinary epidemiologic investigation. Journal ofExposure Science and Environmental Epidemiology , 16 , 191–205.
Alam, M. G. M., Tokunaga, S., & Maekawa, T. (2001). Extraction of arsenic in asynthetic arsenic-contaminated soil using phosphate. Chemosphere, 43 (8), 1035–1041.
Anderson, M. A., Ferguson, J. F., & Gavis, J. (1976). Arsenate adsorption on amorphousaluminum hydroxide. Journal of Colloid and Interface Science, 54 (3), 391–399.
Anderson, M. A., & Malotky, D. (1979). The adsorption of protolyzable anions onhydrous oxides at the isoelectric pH. Journal of Colloid and Interface Science, 72 ,413–427.
Anderson, M. A., Tejedor-Tejedor, M. I., & Stanforth, R. R. (1985). Influence of ag-gregation on the uptake kinetics of phosphate by goethite. Environmental Science &Technology , 19 (7), 632–637.
276
Anderson, P. R., & Benjamin, M. M. (1985). Effect of silicon on the crystallization andadsorption properties of ferric oxides. Environmental Science & Technology , 19 (11),1048–1053.
Anderson, R., Thompson, M., & Culbard, E. (1986). Selective reduction of arsenicspecies by continuous hydride generation. The Analyst , 111 , 1153– 1158.
Appelo, C., Van Der Weiden, M., Tournassat, C., & Charlet, L. (2002). Surface com-plexation of ferrous iron and carbonate on ferrihydrite and the mobilization of arsenic.Environmental Science & Technology , 36 (14), 3096–3103.
Arienzo, M., Adamo, P., Chiarenzelli, J., Bianco, M. R., & De Martino, A. (2002). Re-tention of arsenic on hydrous ferric oxides generated by electrochemical peroxidation.Chemosphere, 48 (10), 1009–1018.
Balasubramanian, N., & Madhavan, K. (2001). Arsenic removal from industrial effluentthrough electrocoagulation. Chemical Engineering & Technology , 24 (5), 519–521.
Bang, S., Johnson, M. D., Korfiatis, G. P., & Meng, X. (2005a). Chemical reactionsbetween arsenic and zero-valent iron in water. Water Research, 39 (5), 763–770.
Bang, S., Korfiatis, G. P., & Meng, X. G. (2005b). Removal of arsenic from water byzero-valent iron. Journal of Hazardous Materials, 121 (1-3), 61–67.
Bard, A., & Faulkner, L. (2001). Electrochemical Methods: Fundamentals and Applica-tions. Wiley and Sons, Inc.
Beck, E. C., Giannini, A. P., & Ramirez, E. R. (1974). Electrocoagulation clarifies foodwastewater. Food Technology , 28 (2), 18.
Beck, R. (2007). World facing arsenic timebomb. BBC News:http://news.bbc.co.uk/2/hi/science/nature/6968574.stm.
Belongia, B. M., Haworth, P. D., Baygents, J. C., & Raghavan, S. (1999). Treatmentof alumina and silica chemical mechanical polishing waste by electrodecantation andelectrocoagulation. Journal of the Electrochemical Society , 146 (11), 4124–4130.
Benjamin, M. M. (1983). Adsorption and surface precipitation of metals on amorphousiron oxyhydroxide. Environmental Science & Technology , 17 (11), 686–692.
Benjamin, M. M. (2002). Water Chemistry . McGraw-Hill.
BGS (2001). Arsenic contamination of groundwater in Bangladesh. Tech. Rep.WC/00/19, British Geological Survey.
Bhattacharya, P., Jacks, G., Ahmed, K. M., Routh, J., & Khan, A. A. (2002). Ar-senic in groundwater of the Bengal Delta Plain aquifers in Bangladesh. Bulletin ofEnvironmental Contamination and Toxicology , 69 (4), 538–545.
277
Birch, W., Pring, A., Reller, A., & Schmalle, H. (1993). Bernalite, Fe(OH)3, a newmineral from Broken Hill, New South Wales: Description and structure. AmericanMineralogist , 78 , 827–834.
Bissen, M., & Frimmel, F. (2003). Arsenic - a review. part II: Oxidation of arsenic andits removal in water treatment. Acta Hydrochimica Et Hydrobiologica, 31 (2), 97–107.
Biswas, N., & Lazarescu, G. (1991). Removal of oil from emulsions using electrocoagu-lation. International Journal of Environmental Studies, 38 , 65.
Bond, A. (2002). Broadening Electrochemical Horizons. New York: Oxford UniversityPress.
Borho, M., & Wilderer, P. (1996). Optimized removal of arsenate(III) by adaptationof oxidation and precipitation processes to the filtration step. Water Science andTechnology , 34 (9), 25–31.
Borho, M., & Wilderer, P. (1997). A reliable method for preservation and detenninationof 138 arsenate(III) concentrations in groundwater and water works samples. Aqua-Journal of Water Supply: Research and Technology , 46 , 138–143.
Boyle, D., Turner, R., & Hall, G. (1998). Anomalous arsenic concentrations in ground-waters of an island community, Bowen Island, British Columbia. Environmental Geo-chemistry and Health, 20 (4), 1573–2983.
BRAC (2000). Combating a deadly menace: Early experience with a community-basedarsenic mitigation project in Bangladesh. Tech. rep., BRAC.
Burgess, W., & Ahmed, K. (2006). Arsenic in aquifers of the Bengal Basin. In R. Naidu,E. Smith, G. Owens, P. Bhattacharya, & P. Nadebaum (Eds.) Managing Arsenic inthe Environment , (pp. 31–58). Australia: CSIRO.
Chang, C.-C., Ho, S.-C., Tsai, S.-S., & Yang, C.-Y. (2004). Ischemic heart disease mor-tality reduction in an arseniasis-endemic area in southwestern Taiwan after a switchin the tap-water supply system. Journal of Toxicology & Environmental Health: PartA, 67 (17), 1353–1361.
Charlet, L., & Manceau, A. (1993). Structure, formation, and reactivity of hydrousoxide particles: insights from x-ray absorption spectroscopy. In J. Buffle, & H. vanLeeuwen (Eds.) Environmental Particles, vol. 2 of Environmental Analytical and Phys-ical Chemistry series. Ann Arbor: Lewis Publishers.
Chen, G. (2004). Electrochemical technologies in wastewater treatment. Separation andPurification Technology , 38 (1), 11–41.
Chen, X. M., Chen, G. H., & Yue, P. L. (2000). Separation of pollutants from restaurantwastewater by electrocoagulation. Separation and Purification Technology , 19 (1-2),65–76.
278
Cheng, R. C., Wang, H. C., & Beuhler, M. D. (1994). Enhanced coagulation for arsenicremoval. Journal of the American Water Works Association, 86 (9), 79–90.
Cherry, J. A., Shaikh, A. U., Tallman, D. E., & Nicholson, R. V. (1979). Arsenic speciesas an indicator of redox conditions in groundwater. Journal of Hydrology , 43 (1-4),373–392.
Chiang, L. C., Chang, J. E., & Tseng, S. C. (1997). Electrochemical oxidation pre-treatment of refractory organic pollutants. Water Science and Technology , 36 (2-3),123–130.
Chiu, H.-F., Ho, S.-C., & Yang, C.-Y. (2004). Lung cancer mortality reduction afterinstallation of tap-water supply system in an arseniasis-endemic area in SouthwesternTaiwan. Lung Cancer , 46 (3), 265–270.
Chowdhury, U. K., Biswas, B. K., Chowdhury, T. R., Samanta, G., Mandal, B. K., Basu,G. C., Chanda, C. R., Lodh, D., Saha, K. C., & Mukherjee, S. K. (2000). Groundwaterarsenic contamination in Bangladesh and West Bengal, India. Environmental HealthPerspectives, 108 (5), 393–397.
Collier, W. (1912). Description of plants at Oklahoma City, Okla and Santa Monica,Calif. Engng Rec., 66 , 55.
Cornell, R., & Schwertmann, U. (2000). The Iron Oxides. Wiley-VCH.
Crisp, P., & Chowdhury, A. (2001). Design of a low-cost purification system for theremoval of arsenic from tubewell water in Bangladesh and India.
Daneshvar, N., Khataee, A. R., Ghadim, A. R. A., & Rasoulifard, M. H. (2007). De-colorization of Ci acid yellow 23 solution by electrocoagulation process: Investigationof operational parameters and evaluation of specific electrical energy consumption(SEEC). Journal of Hazardous Materials, 148 (3), 566–572.
Davies, J. (1995). The hydrogeochemistry of alluvial aquifers in central Bangladesh. InH. Nash, & G. McCall (Eds.) Groundwater Quality , (pp. 9–18). New York: Chapman& Hall.
Davis, C. C., Knocke, W. R., & Edwards, M. (2001). Implications of aqueous silicasorption to iron hydroxide: Mobilization of iron colloids and interference with sorptionof arsenate and humic substances. Environmental Science & Technology , 35 (15), 3158–3162.
Davis, J. A., Fuller, C. C., & Cook, A. D. (1987). A model for trace metal sorptionprocesses at the calcite surface: Adsorption of Cd2+ and subsequent solid solutionformation. Geochimica Et Cosmochimica Acta, 51 (6), 1477–1490.
Davis, J. A., & Kent, D. B. (1990). Surface complexation modeling in aqueous geochem-istry. In M. Hichella, & A. White (Eds.) Mineral-Water Interface Geochemistry , (pp.177–248). Washington D.C.: Mineralogical Society of America.
279
Deng, Y. (1997). Formation of iron(III) hydroxides from homogeneous solutions. WaterResearch, 31 (6), 1347–1354.
Devitre, R., Belzile, N., & Tessier, A. (1991). Speciation and adsorption of arsenic ondiagenetic iron oxyhydroxides. Limnology and Oceanography , 36 (7), 1480–1485.
Dixit, S., & Hering, J. G. (2003). Comparison of arsenic(V) and arsenic(III) sorptiononto iron oxide minerals: Implications for arsenic mobility. Environmental Science &Technology , 37 (18), 4182–4189.
Do, J. S., & Chen, M. L. (1994). Decolourization of dye-containing solutions by electro-coagulation. Journal of Applied Electrochemistry , 24 (8), 785–790.
Driehaus, W., Jekel, M., & Hildebrandt, U. (1998). Granular ferric hydroxide - a newadsorbent for the removal of arsenic from natural water. Journal of Water ServicesResearch and Technology-Aqua, 47 (1), 30–35.
Dzombak, D., & Morel, F. (1990). Surface Complexation Modelling– Hydrous FerricOxide. New York: John Wiley.
Eary, L., & Schramke, J. (1990). Rates of inorganic oxidation reactions involving dis-solved oxygen. In Melchior, & Bassett (Eds.) Chemical Modeling of Aqueous SystemsII . Washington, D.C.: American Chemical Society.
Edwards, M. (1994). Chemistry of arsenic removal during coagulation and Fe-Mn oxi-dation. Journal of the American Water Works Association, 86 (9), 64–78.
Endyuskin, P. N., Selezenkin, S. V., & Dyumaev, K. M. (1983). Electrochemical purifi-cation of wastewaters from production of organic-dyes. Journal of Applied Chemistryof the Ussr , 56 (5), 1100–1102.
EPA, U. (1998). Method 6020A: Inductive coupled plasma-mass spectrometry. Tech.rep., U.S. EPA.
Farquhar, M. L., Charnock, J. M., Livens, F. R., & Vaughan, D. J. (2002). Mechanismsof arsenic uptake from aqueous solution by interaction with goethite, lepidocrocite,mackinawite, and pyrite: An X-ray absorption spectroscopy study. EnvironmentalScience & Technology , 36 (8), 1757–1762.
Farrell, J., Wang, J. P., O’Day, P., & Conklin, M. (2001). Electrochemical and spectro-scopic study of arsenate removal from water using zero-valent iran media. Environ-mental Science & Technology , 35 (10), 2026–2032.
Fendorf, S., Eick, M. J., Grossl, P., & Sparks, D. L. (1997). Arsenate and chromateretention mechanisms on goethite .1. surface structure. Environmental Science &Technology , 31 (2), 315–320.
280
Ferguson, J. F., & Anderson, M. A. (1974). Chemical form of arsenic in water suppliesand their removal. In A. Rubin (Ed.) Chemistry of Water Supply, Treatment, andDistribution, (pp. 137–158). Ann Arbor, MI: Ann Arbor Science.
Ferguson, J. F., & Gavis, J. (1972). Review of arsenic cycle in natural waters. WaterResearch, 6 (11), 1259–1274.
Ford, R. G. (2002). Rates of hydrous ferric oxide crystallization and the influence oncoprecipitated arsenate. Environmental Science & Technology , 36 (11), 2459–2463.
Foster, A. (1999). Partitioning and Transformation of Arsenic and Selenium in Naturaland Laboratory Systems. Ph.D. thesis, Stanford University.
Foster, A. L. (2003). Spectroscopic investigations of arsenic species in solid phases.In A. H. Welch, & K. Stollenwerk (Eds.) Arsenic in Groundwater . Boston: KluwerAcademic Press.
Fuller, C. C., Davis, J. A., & Waychunas, G. A. (1993). Surface-chemistry of ferrihydrite.2. kinetics of arsenate adsorption and coprecipitation. Geochimica Et CosmochimicaActa, 57 (10), 2271–2282.
Gallagher, P. A., Schwegel, C. A., Parks, A., Gamble, B. M., Wymer, L., & Creed,J. T. (2004). Preservation of As(III) and As(V) in drinking water supply samplesfrom across the United States using EDTA and acetic acid as a means of minimizingiron-arsenic coprecipitation. Environmental Science & Technology , 38 (10), 2919–2927.
Gao, Y., & Mucci, A. (2001). Acid base reactions, phosphate and arsenate complexation,and their competitive adsorption at the surface of goethite in 0.7 M NaCl solution.Geochimica Et Cosmochimica Acta, 65 (14), 2361–2378.
Ghosh, M., & Teoh, R. (1985). Adsorption of arsenic on hydrous aluminum oxide.In Toxic and Hazardous Wastes: Proceedings of the Seventh Mid-Atlantic IndustrialWaste Conference June 23-25, 1985., (pp. 139–155). Lancaster, PA.: Technomic Pub-lishing Co.
Goldberg, S. (1986). Chemical modeling of arsenate adsorption on aluminum and iron-oxide minerals. Soil Science Society of America Journal , 50 (5), 1154–1157.
Goldberg, S., & Johnston, C. T. (2001). Mechanisms of arsenic adsorption on amor-phous oxides evaluated using macroscopic measurements, vibrational spectroscopy,and surface complexation modeling. Journal of Colloid and Interface Science, 234 (1),204–216.
Gomes, J. A. G., Daida, P., Kesmez, M., Weir, M., Moreno, H., Parga, J. R., Irwin,G., McWhinney, H., Grady, T., Peterson, E., & Cocke, D. L. (2007). Arsenic removalby electrocoagulation using combined Al-Fe electrode system and characterization ofproducts. Journal of Hazardous Materials, 139 (2), 220–231.
281
Grossl, P. R., Eick, M., Sparks, D. L., Goldberg, S., & Ainsworth, C. C. (1997). Arsenateand chromate retention mechanisms on goethite .2. kinetic evaluation using a pressure-jump relaxation technique. Environmental Science & Technology , 31 (2), 321–326.
Grossl, P. R., & Sparks, D. L. (1995). Evaluation of contaminant ion adsorp-tion/desorption on goethite using pressure jump relaxation kinetics. Geoderma, 67 (1-2), 87–101.
Gulledge, J. H., & Oconnor, J. T. (1973). Removal of arsenic (V) from water by ad-sorption on aluminum and ferric hydroxides. Journal of the American Water WorksAssociation, 65 (8), 548–552.
Gupta, S. K., & Chen, K. Y. (1978). Arsenic removal by adsorption. Journal WaterPollution Control Federation, 50 (3), 493–506.
Gurses, A., Yalcin, M., & Dogar, C. (2002). Electrocoagulation of some reactive dyes: astatistical investigation of some electrochemical variables. Waste Management , 22 (5),491–499.
Hansen, H. K., Nunez, P., & Grandon, R. (2006). Electrocoagulation as a remediationtool for wastewaters containing arsenic. Minerals Engineering , 19 (5), 521–524.
Hansen, H. K., Nunez, P., Raboy, D., Schippacasse, I., & Grandon, R. (2007). Elec-trocoagulation in wastewater containing arsenic: Comparing different process designs.Electrochimica Acta, 52 (10), 3464–3470.
Harvey, C. F., Swartz, C. H., Badruzzaman, A. B. M., Keon-Blute, N., Yu, W., Ali,M. A., Jay, J., Beckie, R., Niedan, V., Brabander, D., Oates, P. M., Ashfaque, K. N.,Islam, S., Hemond, H. F., & Ahmed, M. F. (2002). Arsenic mobility and groundwaterextraction in Bangladesh. Science, 298 (5598), 1602–1606.
Hem, J. (1961). Stability field diagrams as aids in iron chemistry studies. Journal of theAmerican Water Works Association, 53 (2), 211–232.
Hering, J. G., Chen, P. Y., Wilkie, J. A., Elimelech, M., & Liang, S. (1996). Arsenicremoval by ferric chloride. Journal of the American Water Works Association, 88 (4),155–167.
Hiemstra, T., & Van Riemsdijk, W. H. (1999). Surface structural ion adsorption modelingof competitive binding of oxyanions by metal (hydr)oxides. Journal of Colloid andInterface Science, 210 (1), 182–193.
Hingston, F. J., Posner, A. M., & Quirk, J. P. (1971). Competitive adsorption of neg-atively charged ligands on oxide surfaces. Discussions of the Faraday Society , 52 ,334–342.
Holm, T. (2002). Effects of co32-/bicarbonate, si, and po43- on arsenic sorption to HFO.Journal of the American Water Works Association, 94 (4), 174–181.
282
Holm, T., Anderson, M. A., Iverson, D., & Stanforth, R. (1979). Heterogeneous of arsenicin aquatic systems. American Chemical Society Symposium Series, 93 , 711–736.
Holt, P. K., Barton, G. W., Wark, M., & Mitchell, C. A. (2002). A quantitative com-parison between chemical dosing and electrocoagulation. Colloids and Surfaces A-Physicochemical and Engineering Aspects, 211 (2-3), 233–248.
Honeyman, B. D., & Santschi, P. H. (1991). Coupling adsorption and particle aggrega-tion: Laboratory studies of colloidal pumping‘ using 59-Fe-labeled hematite. Environ-mental Science & Technology , 25 , Pages: 1739–1747.
Hopenhayn-Rich, C., Biggs, M., Kalman, D., Moore, L., & Smith, A. (1996). Arsenicmethylation patterns before and after changing from high to lower concentrations ofarsenic in drinking water. Environmental Health Perspectives, 104 (11), 1200–1207.
Hsia, T. H., Lo, S. L., & Lin, C. F. (1992). As(V) adsorption on amorphous iron-oxide- triple layer modeling. Chemosphere, 25 (12), 1825–1837.
Hsia, T. H., Lo, S. L., Lin, C. F., & Lee, D. (1994). Characterization of arsenate adsorp-tion on hydrous iron oxide using chemical and physical methods. Characterizationof arsenate adsorption on hydrous iron oxide using chemical and physical methods,85 (11), 1–7.
Hug, S. J., & Leupin, O. (2003). Iron-catalyzed oxidation of arsenic(III) by oxygen andby hydrogen peroxide; ph-dependent formation of oxidants in the Fenton reaction.Environmental Science & Technology , 37 (12), 2734–2742.
Ibanez, J. G., Singh, M. M., & Szafran, Z. (1998). Laboratory experiments on electro-chemical remediation of the environment. Part 4: Color removal of simulated wastew-ater by electrocoagulation-electroflotation. Journal of Chemical Education, 75 (8),1040–1041.
Islam, F. S., Boothman, C., Gault, A. G., Polya, D. A., & Lloyd, J. R. (2005). Poten-tial role of the Fe(III)-reducing bacteria geobacter and geothrix in controlling arsenicsolubility in Bengal delta sediments. Mineralogical Magazine, 69 (5), 865–875.
Islam, F. S., Gault, A. G., Boothman, C., Polya, D. A., Charnock, J. M., Chatterjee, D.,& Lloyd, J. R. (2004). Role of metal-reducing bacteria in arsenic release from Bengaldelta sediments. Nature, 430 (6995), 68–71.
Jackson, B. P., & Miller, W. P. (2000). Effectiveness of phosphate and hydroxide fordesorption of arsenic and selenium species from iron oxides. Soil Science Society ofAmerica Journal , 64 (5), 1616–1622.
Jain, A., & Loeppert, R. H. (2000). Effect of competing anions on the adsorption ofarsenate and arsenite by ferrihydrite. Journal of Environmental Quality , 29 (5), 1422–1430.
283
Jain, A., Raven, K. P., & Loeppert, R. H. (1999). Arsenite and arsenate adsorption onferrihydrite: Surface charge reduction and net OH- release stoichiometry. Environ-mental Science & Technology , 33 (8), 1179–1184.
Jovanovic, V., & Hackerman, N. (1998). Electrochemical passivation of iron in NO3-,SO42-, and ClO4- solutions. Journal of Physical Chemistry B , 102 (49), 9855–9860.
Kahn, M. (2007). Arsenic in water a risk to 140 million people. Reuters:http://www.alertnet.org/thenews/newsdesk/L29757483.htm.
Kaufmann, R., Sorensen, B., Rahman, M., Streatfield, K., & Persson, L. (2001). Ad-dressing the public health crisis caused by arsenic contamination of drinking water inBangladesh. Tech. rep., South Asia Health, Nutrition, and population Unit, WorldBank.
Kim, M.-J., & Nriagu, J. (2000). Oxidation of arsenite in groundwater using ozone andoxygen. The Science of The Total Environment , 247 (1), 71–79.
Kim, M. J., Nriagu, J., & Haack, S. (2000). Carbonate ions and arsenic dissolution bygroundwater. Environmental Science & Technology , 34 (15), 3094–3100.
Kolitsch, U. (1998). Bernalite from Clara Mine, Germany, and the incorporation oftungsten in minerals containing ferric iron. Canad. Min., 36 , 1211–1216.
Koparal, A. S., & Ogutveren, U. B. (2002). Removal of nitrate from water by electrore-duction and electrocoagulation. Journal of Hazardous Materials, 89 (1), 83–94.
Korte, N. E., & Fernando, Q. (1991). A review of arsenic(iii) in groundwater. CriticalReviews in Environmental Control , 21 (1), 1–39.
Kumar, P. R., Chaudhari, S., Khilar, K. C., & Mahajan, S. P. (2004). Removal of arsenicfrom water by electrocoagulation. Chemosphere, 55 (9), 1245–1252.
Kurbatov, M. H., & Wood, G. B. (1952). Rate of adsorption of cobalt ions on hydrousferric oxide. The Journal of Physical Chemistry , 56 (6), 698–701.
Kyle, M., & Christiani, D. (2008). Environmental arsenic exposure and diabetes. Journalof the Americal Medical Association, 300 (7), 845–846.
Lackovic, J. A., Nikolaidis, N. P., & Dobbs, G. M. (2000). Inorganic arsenic removal byzero-valent iron. Environmental Engineering Science, 17 (1), 29–39.
Leckie, J. O., Benjamin, M. M., Hayes, K., Kaufman, G., & Altmann, S. (1980). Ad-sorption/coprecipitation of trace elements from water with iron oxyhydroxide. Tech.rep., Stanford Univ., CA (USA). Dept. of Civil Engineering.
Liu, F., De Cristofaro, A., & Violante, A. (2001). Effect of pH, phosphate and oxalateon the adsorption/desorption of arsenate on/from goethite. Soil Science, 166 (3), 197–208.
284
Lowers, H. A., Breit, G. N., Foster, A. L., Whitney, J., Yount, J., Uddin, M. N., &Muneem, A. A. (2007). Arsenic incorporation into authigenic pyrite, Bengal Basinsediment, Bangladesh. Geochimica Et Cosmochimica Acta, 71 (11), 2699–2717.
Lytle, D., & Snoeyink, V. (2002). Effect of ortho- and polyphosphates on the propertiesof iron particles and suspensions. Journal of the American Water Works Association,94 (10), 87–99.
Madajewicz, M., Pfaff, A., van Geen, A., Graziano, J., Hussein, I., Momotaj, H., Sylvi,R., & Ahsan, H. (2008). Can information alone change behavior? Response to ar-senic contamination of groundwater in Bangladesh (vol 84, pg 731, 2007). Journal ofDevelopment Economics, 85 (1-2), 349–349.
Mameri, N., Yeddou, A. R., Lounici, H., Belhocine, D., Grib, H., & Bariou, B. (1998).Defluoridation of septentrional Sahara water of north Africa by electrocoagulationprocess using bipolar aluminium electrodes. Water Research, 32 (5), 1604–1612.
Mamtaz, R., & Bache, D. (2001). Reduction of arsenic in groundwater by coprecipitationwith iron. Journal of Water Supply: Research & Technology- AQUA, 50 , 313–324.
Manceau, A. (1995). The mechanism of anion adsorption on iron-oxides - evidencefor the bonding of arsenate tetrahedra on free Fe(O,OH)(6) edges. Geochimica EtCosmochimica Acta, 59 (17), 3647–3653.
Manceau, A., & Combes, J. M. (1988). Structure of Mn and Fe oxides and oxyhydroxides- a topological approach by EXAFS. Physics and Chemistry of Minerals, 15 (3), 283–295.
Manning, B. A., Fendorf, S. E., & Goldberg, S. (1998). Surface structures and sta-bility of arsenic(iii) on goethite: Spectroscopic evidence for inner-sphere complexes.Environmental Science & Technology , 32 (16), 2383–2388.
Manning, B. A., & Goldberg, S. (1996a). Modeling arsenate competitive adsorption onkaolinite, montmorillonite and illite. Clays and Clay Minerals, 44 (5), 609–623.
Manning, B. A., & Goldberg, S. (1996b). Modeling competitive adsorption of arsenatewith phosphate and molybdate on oxide minerals. Soil Science Society of AmericaJournal , 60 (1), 121–131.
Marcus, M., MacDowell, A., R., C., Manceau, A., Miller, T., Padmore, H., & Sublett,R. (2004). Beamline 10.3.2 at ALS: a hard x-ray microprobe for environmental andmaterials sciences. Journal of Synchrotron Radiation, 11 , 239–247.
Martin, T. A., & Kempton, J. H. (2000). In situ stabilization of metal-contaminatedgroundwater by Hydrous Ferric Oxide: An experimental and modeling investigation.Environmental Science & Technology , 34 (15), 3229–3234.
285
Mathieu, J., Gadgil, A., Kowolik, K., & Addy, S. (2008). Removing arsenic from con-taminated drinking water in rural Bangladesh: recent fieldwork results and policyimplications. In Safe Drinking Water: Where Science Meets Policy . Chapel Hill, NC.
Matteson, M. J., Dobson, R. L., Glenn, R. W., Kukunoor, N. S., Waits, W. H., &Clayfield, E. J. (1995). Electrocoagulation and separation of aqueous suspensions ofultrafine particles. Colloids and Surfaces A-Physicochemical and Engineering Aspects,104 (1), 101–109.
Mayer, T. D., & Jarrell, W. M. (2000). Phosphorus sorption during iron(II) oxidationin the presence of dissolved silica. Water Research, 34 (16), 3949–3956.
Melitas, N., Conklin, M., & Farrell, J. (2002a). Electrochemical study of arsenate andwater reduction on iron media used for arsenic removal from potable water. Environ-mental Science & Technology , 36 (14), 3188–3193.
Melitas, N., Wang, J. P., Conklin, M., O’Day, P., & Farrell, J. (2002b). Understandingsoluble arsenate removal kinetics by zerovalent iron media. Environmental Science &Technology , 36 (9), 2074–2081.
Meng, X., Bang, S., & Korfiatis, G. P. (2000). Effects of silicate, sulfate, and carbonateon arsenic removal by ferric chloride. Water Research, 34 (4), 1255–1261.
Meng, X., Korfiatis, G. P., Jing, C., & Christodoulatos, C. (2001). Redox transformationsof arsenic and iron in water treatment sludge during aging and TCLP extraction.Environmental Science & Technology , 35 (17), 3476–3481.
Meng, X., & Wang, W. (1998). Speciation of arsenic by disposable cartridges. In Bookof posters of the third International Conference on arsenic exposure and health effects:Society of Environmental Geochemistry and Health. Denver: University of Colorado -Denver.
Meng, X. G., Korfiatis, G. P., Bang, S. B., & Bang, K. W. (2002). Combined effects ofanions on arsenic removal by iron hydroxides. Toxicology Letters, 133 (1), 103–111.
Misawa, T. (1973). Thermodynamic consideration for Fe-H2O system at 25 degrees C.Corrosion Science, 13 (9), 659–676.
Mishra, D. (2006). Electrochemical Deactivation of Nitrate, Arsenate, and Trichloroethy-lene. Ph.D. thesis, University of Arizona.
Mishra, D., & Farrell, J. (2005). Evaluation of mixed valent iron oxides as reactiveadsorbents for arsenic removal. Environmental Science & Technology , 39 (24), 9689–9694.
Mollah, M. Y. A., Morkovsky, P., Gomes, J. A. G., Kesmez, M., Parga, J., & Cocke,D. L. (2004). Fundamentals, present and future perspectives of electrocoagulation.Journal of Hazardous Materials, 114 (1-3), 199–210.
286
Moore, L. E., Smith, A. H., Hopenhayn-Rich, C., Biggs, M. L., Kalman, D. A., & Smith,M. T. (1997). Decrease in bladder cell micronucleus prevalence after intervention tolower the concentration of arsenic in drinking water. Cancer Epidemiology Biomarkers& Prevention, 6 (12), 1051–1056.
Navas-Acien, A., Silbergeld, E., Pastor-Barriuso, R., & Guallar, E. (2008). Arsenicexposure and prevalence of type 2 diabetes in US adults. Journal of the AmericanMedical Association, 300 (7), 814–822.
Ng, K., Ujang, Z., & Le-Clech, P. (2004). Arsenic removal technologies for drinkingwater treatment. Reviews in Environmental Science and Bio/Technology , 3 , 43–53.
Nickson, R. T., McArthur, J. M., Ravenscroft, P., Burgess, W. G., & Ahmed, K. M.(2000). Mechanism of arsenic release to groundwater, Bangladesh and West Bengal.Applied Geochemistry , 15 (4), 403–413.
O’Day, P. A., Rehr, J. J., Zabinsky, S. I., & Brown, J., G. E. (1994). Extended X-rayAbsorption Fine Structure (EXAFS) analysis of disorder and multiple-scattering incomplex crystalline solids. Journal of the American Chemical Society , 116 (7), 2938–2949.
O’Melia, C. (1987). Particle-particle interactions. In W. Stumm (Ed.) Aquatic SurfaceChemistry , (pp. 385–403). J. Wiley.
O’Reilly, S. E., Strawn, D. G., & Sparks, D. L. (2001). Residence time effects on arse-nate adsorption/desorption mechanisms on goethite. Soil Science Society of AmericaJournal , 65 (1), 67–77.
Paktunc, D., Foster, A., & Laflamme, G. (2003). Speciation and characterization of ar-senic in Ketza River mine tailings using X-ray absorption spectroscopy. EnvironmentalScience & Technology , 37 (10), 2067–2074.
Panizza, M., Bocca, C., & Cerisola, G. (2000). Electrochemical treatment of wastewatercontaining polyaromatic organic pollutants. Water Research, 34 (9), 2601–2605.
Parga, J. R., Cocke, D. L., Valenzuela, J. L., Gomes, J. A., Kesmez, M., Irwin, G.,Moreno, H., & Weir, M. (2005a). Arsenic removal via electrocoagulation from heavymetal contaminated groundwater in La Comarca Lagunera Mexico. Journal of Haz-ardous Materials, 124 (1-3), 247–254.
Parga, J. R., Cocke, D. L., Valverde, V., Gomes, J. A. G., Kesmez, M., Moreno, H.,Weir, M., & Mencer, D. (2005b). Characterization of electrocoagulation for removalof chromium and arsenic. Chemical Engineering & Technology , 28 (5), 605–612.
Persin, F., & Rumeau, M. (1989). Electrochemical treatment of waters and effluents.Tribune de l’eau, 42 (3), 45–66.
287
Petrusevski, B., Boere, J., Shahidullah, S. M., Sharma, S. K., & Schippers, J. C. (2002).Adsorbent-based point-of-use system for arsenic removal in rural areas. Aqua- Journalof Water Supply: Research and Technology , 51 (3), 135–144.
Petrusevski, B., Sharma, S., Schippers, J. C., & Shordt, K. (2007). Arsenic in drinkingwater. Tech. rep., IRC International Water and Sanitation Centre.
Phutdhawong, W., Chowwanapoonpohn, S., & Buddhasukh, D. (2000). Electrocoagu-lation and subsequent recovery of phenolic compounds. Analytical Sciences, 16 (10),1083–1084.
Pierce, M. L., & Moore, C. B. (1982). Adsorption of arsenite and arsenate on amorphousiron hydroxide. Water Research, 16 (7), 1247–1253.
Pinisakul, A., Polprasert, C., Parkpian, P., & Satayavivad, J. (2002). Arsenic removalefficiency and mechanisms by electro-chemical precipitation process. Water Scienceand Technology , 46 (9), 247–254.
Pogrebnaya, V. L., Klimenko, A. A., Bokovikova, T. N., Tsymbal, E. P., & Pronina, N. P.(1995). Purification of waste water of heavy metals by electrocoagulation. Chemicaland Petroleum Engineering , 31 (5), 280–281.
Posselt, H. S., Anderson, F. J., & Weber, W. J. (1968). Cation sorption on colloidalhydrous manganese dioxide. Environmental Science & Technology , 2 (12), 1087–1093.
Pouet, M. F., & Grasmick, A. (1995). Urban waste-water treatment by electrocoagulationand flotation. Water Science and Technology , 31 (3-4), 275–283.
Rahman, M. M., Sengupta, M. K., Ahamed, S., Chowdhury, U. K., Hossain, M., Das,B., Lodh, D., Saha, K. C., Pati, S., & Kaies, I. (2005). The magnitude of arseniccontamination in groundwater and its health effects to the inhabitants of the Jalangi;one of the 85 arsenic affected blocks in West Bengal, India. Science of the TotalEnvironment , 338 (3), 189–200.
Raven, K. P., Jain, A., & Loeppert, R. H. (1998). Arsenite and arsenate adsorption onferrihydrite: Kinetics, equilibrium, and adsorption envelopes. Environmental Science& Technology , 32 (3), 344–349.
Ravenscroft, P. (2003). An overview of the hydrogeology of Bangladesh. In A. Rahman, &P. Ravenscroft (Eds.) Groundwater Resources and Development in Bangladesh. Dhaka:Bangladesh Center for Advanced Studies: University Press.
Rehr, J. J., Deleon, J. M., Zabinsky, S. I., & Albers, R. C. (1991). Theoretical x-ray absorption fine-structure standards. Journal of the American Chemical Society ,113 (14), 5135–5140.
Roberts, L. C., Hug, S. J., Ruettimann, T., Billah, M. M., Khan, A. W., & Rahman,M. T. (2004). Arsenic removal with iron(II) and iron(III) in waters with high silicateand phosphate concentrations. Environmental Science & Technology , 38 (1), 307–315.
288
Roman-Ross, G., Cuello, G. J., Turrillas, X., Fernandez-Martinez, A., & Charlet, L.(2006). Arsenite sorption and co-precipitation with calcite. Chemical Geology , 233 (3-4), 328–336.
Roy, J. (2008). Economic benefits of arsenic removal from ground water – a case studyfrom West Bengal, India. Science of the Total Environment , 397 (1-3), 1–12.
Rushing, J. C., McNeill, L. S., & Edwards, M. (2003). Some effects of aqueous silica onthe corrosion of iron. Water Research, 37 (5), 1080–1090.
Ryden, J. C., Syers, J. K., & Tillman, R. W. (1987). Inorganic anion sorption andinteractions with phosphate sorption by hydrous ferric oxide gel. European Journal ofSoil Science, 38 (2), 211–217.
Saha, J. C., Dikshit, A. K., Bandyopadhyay, M., & Saha, K. C. (1999). A review ofarsenic poisoning and its effects on human health. Critical Reviews in EnvironmentalScience and Technology , 29 (3), 281–313.
Saha, K. (1995). Chronic arsenical dermatoses from tube-well water in West Bengalduring 1983-87. Indian Journal of Dermatology , 40 , 1–12.
Sato, M., Kimura, M., Yamashita, M., Konishi, H., Fujimoto, S., Tabira, Y., Doi, T.,Nagoshi, M., Suzuki, S., Kamimura, T., Nakayama, T., & Ohtsuka, T. (2005). Atomic-structure characterization of passive film of fe by grazing incidence x-ray scatteringat SPring-8. In P. Marcus, & V. Maurice (Eds.) Passivation of metals and Semicon-ductors, and Properties of Thin Oxide Layers - 9th International Symposium. Paris,France: Elsevier.
Schenk, J., & Weber, W. (1968). Chemical interactions of dissolved silica with iron(II)and iron(III). Journal of the American Water Works Association, 60 , 199–212.
Shen, Y. S. (1973). Study of arsenic removal from drinking-water. Journal of the Amer-ican Water Works Association, 65 (8), 543–548.
Sigg, L., & Stumm, W. (1981). The interaction of anions and weak acids with thehydrous goethite ([alpha]-FeOOH) surface. Colloids and Surfaces, 2 (2), 101–117.
Smedley, P. L., & Kinniburgh, D. G. (2002). A review of the source, behaviour anddistribution of arsenic in natural waters. Applied Geochemistry , 17 (5), 517–568.
Smith, A. H., Lingas, E. O., & Rahman, M. (2000). Contamination of drinking-waterby arsenic in Bangladesh: a public health emergency. Bulletin of the World HealthOrganization, 78 (9).
Sorg, T. J., & Logsdon, G. S. (1978). Treatment technology to meet the interim primarydrinking water regulations for inorganics: Part 2. Journal of the American WaterWorks Association, 70 (7), 379–393.
289
Sposito, G. A. (1986). On distiguishing adsorption from surface precipitation. InJ. Davies, & K. Hayes (Eds.) Geochemical Processes at Mineral Surfaces, (pp. 217–228). American Chemical Society Symposium Series.
Stollenwerk, K. (2003). Geochemical processes controlling transport of arsenic in ground-water: A review of adsorption. In A. H. Welch, & K. Stollenwerk (Eds.) Arsenic inGroundwater . Boston: Kluwer Academic Publishers.
Stollenwerk, K., & Kipp, K. (1990). Simulation of molybdate transport with differentrate-controlled mechanisms. In D. Melchior, & R. Bassett (Eds.) Chemical Model-ing of Aqueous Systems II: Washington, (pp. 243–257). American Chemical SocietySymposium Series.
Stumm, W., & Morgan, J. (1996). Aquatic Chemistry . John Wiley & Sons, Inc.
Su, C. M., & Puls, R. W. (2001). Arsenate and arsenite removal by zerovalent iron:Kinetics, redox transformation, and implications for in situ groundwater remediation.Environmental Science & Technology , 35 (7), 1487–1492.
Sun, X. H., & Doner, H. E. (1996). An investigation of arsenate and arsenite bondingstructures on goethite by FTIR. Soil Science, 161 (12), 865–872.
Swedlund, P. J., & Webster, J. G. (1999). Adsorption and polymerisation of silicicacid on ferrihydrite, and its effect on arsenic adsorption. Water Research, 33 (16),3413–3422.
Szynkarczuk, J., Kan, J., Hassan, T. A. T., & Donini, J. C. (1994). Electrochemicalcoagulation of clay suspensions. Clays and Clay Minerals, 42 (6), 667–673.
Tang, Y. K., & Furtak, T. E. (1991). Study of underpotential deposition of metals usingthe quartz crystal microbalance. Electrochimica Acta, 36 (11-12), 1869–1872.
Teo, B. K., & Joy, D. (1981). EXAFS Spectroscopy, Techniques and Applications. NewYork, NY: Plenum Press.
Toner, B. M., Santelli, C. M., Marcus, M. A., Wirth, R., Chan, C. S., McCollom, T. M.,Bach, W., & Edwards, K. J. (2008). Biogenic iron oxide formation at Mid-OceanRidge hydrothermal vents: Juan de Fuca Ridge. Geochimica et Cosmochimica Acta(accepted).
Tsai, C. T., Lin, S. T., Shue, Y. C., & Su, P. L. (1997). Electrolysis of soluble organicmatter in leachate from landfills. Water Research, 31 (12), 3073–3081.
UNICEF (1999). Progotir Pathey (on the road to progress); acheiving the goals forchildren in Bangladesh. Tech. rep., UNICEF.
UNICEF (2006). Arsenic mitigation in Bangladesh fact sheet. Tech. rep., UNICEF.
290
van Green, A., Robertson, A. P., & Leckie, J. O. (1994). Complexation of carbonatespecies at the goethite surface: Implications for adsorption of metal ions in naturalwaters. Geochimica Et Cosmochimica Acta, 58 (9), 2073–2086.
van Riemsdijk, W. H., Boumans, L. J. M., & de Haan, F. A. M. (1984). Phosphatesorption by soils: I. a model for phosphate reaction with metal-oxides in soil. SoilScience Society of America Journal , 48 (3), 537–541.
Vik, E. A., Carlson, D. A., Eikum, A. S., & Gjessing, E. T. (1984). Electrocoagulationof potable water. Water Research, 18 (11), 1355–1360.
Villalobos, M., & Leckie, J. O. (2000). Carbonate adsorption on goethite under closedand open CO2 conditions. Geochimica Et Cosmochimica Acta, 64 (22), 3787–3802.
Vlyssides, A. G., Loizidou, M., Karlis, P. K., Zorpas, A. A., & Papaioannou, D. (1999).Electrochemical oxidation of a textile dye wastewater using a Pt/Ti electrode. Journalof Hazardous Materials, 70 (1-2), 41–52.
Vlyssides, A. G., Papaioannou, D., Loizidoy, M., Karlis, P. K., & Zorpas, A. A. (2000).Testing an electrochemical method for treatment of textile dye wastewater. WasteManagement , 20 (7), 569–574.
Waltham, C. A., & Eick, M. J. (2002). Kinetics of arsenic adsorption on goethite inthe presence of sorbed silicic acid. Soil Science Society of America Journal , 66 (3),818–825.
Wasserman, G., Liu, X., Parvez, F., Ahsan, H., Factor-Litvak, P., van Geen, A.,Slavkovich, V., LoIacono, N., Cheng, Z., Hussain, I., Momotaj, H., & J.H., G. (2004).Water arsenic exposure and children’s intellectual function in Araihazar, Bangladesh.Environmental Health Perspectives, 112 (13), 1329–1333.
Waychunas, G. A., Rea, B. A., Fuller, C. C., & Davis, J. A. (1993). Surface chemistryof ferrihydrite. 1. EXAFS studies of the geometry of coprecipitated and adsorbedarsenate. Geochimica Et Cosmochimica Acta, 57 (10), 2251–2269.
Webb, S. (2003). SIXPack: A graphical interface for XAS analysis using IFFEFFITpresented by Samuel Webb. Physica Scripta, T115 .
Welch, A. H., Lico, M. S., & Hughes, J. L. (1988). Arsenic in ground-water of the westernUnited States. Ground Water , 26 (3), 333–347.
WHO (1993). Guidlines for drinking-water quality. Tech. Rep. ISBN 92 4 154460, TheWorld Health Organization.
WHO (2000). Towards an assessment of the socioeconomic impact of arsenic poisoningin Bangladesh. Tech. rep., World Health Organization.
WHO (2004). Guidelines for drinking water quality. Vol 1. recommendations. Tech. rep.,World Health Organization.
291
Wilkie, J. A., & Hering, J. G. (1996). Adsorption of arsenic onto hydrous ferric oxide:effects of adsorbate/adsorbent ratios and co-occurring solutes. Colloids and SurfacesA: Physicochemical and Engineering Aspects, 107 , 97–110.
Willett, I. R., Chartres, C. J., & Nguyen, T. T. (1988). Migration of phosphate intoaggregated particles of ferrihydrite. European Journal of Soil Science, 39 (2), 275–282.
WSP-SA (2000). Arsenic mitigation in West Bengal and Bangladesh: helping householdsrespond to a water quality crisis. Tech. rep., Water and Sanitation Program - SouthAsia.
Wu, S. C., & Gschwend, P. M. (1988). Numerical modeling of sorption kinetics oforganic compounds to soil and sediment particles. Water Resources Research, 24 ,Pages: 1373–1383.
Xiong, Y., Strunk, P. J., Xia, H. Y., Zhu, X. H., & Karlsson, H. T. (2001). Treat-ment of dye wastewater containing acid orange ii using a cell with three-phase three-dimensional electrode. Water Research, 35 (17), 4226–4230.
Yamauchi, H., & Fowler, B. (1994). Toxicity and metabolism of inorganic and methy-lated arsenicals. In J. Nriagu (Ed.) Arsenic in the Environment, Part I. Cycling andCharacterization, (p. 35). New York: Wiley-Interscience.
Yildiz, Y. S., Koparal, A. S., Irdemez, S., & Keskinler, B. (2007). Electrocoagulation ofsynthetically prepared waters containing high concentration of NOM using iron castelectrodes. Journal of Hazardous Materials, 139 (2), 373–380.
Zachara, J. M., Girvin, D. C., Schmidt, R. L., & Resch, C. T. (1987). Chromate ad-sorption on amorphous iron oxyhydroxide in the presence of major groundwater ions.Environmental Science & Technology , 21 (6), 589–594.
292
Appendix A
List of Acronyms
• ALS: Advanced Light Source (Berkeley, CA).
• EC: ElectroCoagulation.
• ECAR: ElectroChemical Arsenic Remediation.
• EGA: ECAR-Generated Adsorbent (refers to the iron (hydr)oxide sludge gen-erated by iron dissolution and subsequent hydrolysis during the dosing stage ofECAR).
• EXAFS: Extended X-ray Absorption Fine Structure.
• FT-IR: Fourier Transform Infrared spectroscopy.
• HFO: Hydrous Ferric Oxide (a.k.a. Ferrihydrite, am-Fe(OH)3).
• PSA: Post-Synthesis Adsorption.
• PS-EGA: Post-Synthesis ECAR-Generated iron (hydr)oxide Adsorbent (i.e. EGAformed in the absence of arsenic).
• SBGW: Synthetic Bangladesh groundwater.
• SBGW-1: Synthetic Bangladesh groundwater using recipe 1, with low phosphate,no silicate, and high sulfate relative to Bangladesh averages; composition describedin Table 3.1.
• SBGW-2: Synthetic Bangladesh groundwater using recipe 2, with average levelsof phosphate, silicate, and sulfate relative to Bangladesh; composition described inTable 3.1.
• SEM: Scanning Electron Microscopy.
• SCE: Saturated Calomel Electrode.
• SHE: Standard Hydrogen Electrode.
293
• SSRL: Stanford Synchrotron Radiation Laboratory (San Jose, CA).
• XANES: X-ray Absorption Near Edge Structure.
• XRF: X-ray Fluorescence (as in XRF maps or XRF mapping).
• XAS: X-ray Absorption Spectroscopy.
• XPS: X-ray Photon Spectroscopy.
• XRD: X-ray Diffraction.
• ZVI: Zero-Valent Iron.